Production Preprocess
Xinyu Zhang
2023-03-27
Measurements from genders and phonemes
library(dplyr)##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionfi<- read.csv("allFemaleI.csv", header = TRUE, sep = "\t")
fig<- fi |> mutate(gender = c(rep("F", times = nrow(fi))))
fe <- read.csv("allFemaleE.csv", header = TRUE, sep = "\t")
feg<- fe |> mutate(gender = c(rep("F", times = nrow(fe))))
mi<- read.csv("allMaleI.csv", header = TRUE, sep = "\t")
mig<- mi |> mutate(gender = c(rep("M", times = nrow(mi))))
me <- read.csv("allMaleE.csv", header = TRUE, sep = "\t")
meg<- me |> mutate(gender = c(rep("M", times = nrow(me))))
mfie<- rbind(fig, feg, mig, meg)
#extract the fileName column
#fileNames<- rcds |> select(fileName)
#head(fileNames)splitting the different segments in the file name into columns because they provide info about the recording
prod <- read.table(text = as.character(mfie$fileName), sep = "_")
head(prod)## V1 V2 V3 V4 V5
## 1 subj 10 kiggif 13 Post
## 2 subj 10 kiggif 17 Post
## 3 subj 10 kiggif 26 Post
## 4 subj 10 kiggif AF2 28
## 5 subj 10 kiggif AF2 52
## 6 subj 10 kiggif MN1 1### removing the ".wav" from the final column
#```{r}
#library(stringr)
#library(dplyr)
#prod <- prod %>%
# mutate(V5 =
# str_sub(prod$V5, end=-5)
# )
#head(prod)renaming the column names accordingly
#removing column 1 because it's just "subj"
newcols <- prod %>%
select(-V1)
colnames(newcols)## [1] "V2" "V3" "V4" "V5"#revert the "Post" and the "seq" for posttest items
#if v5 = post, v5 = v4, v4 = v5
correctPost <- newcols |>
mutate(V51 = ifelse(V5 == "Post", V4, V5),
V4 = ifelse(V5 == "Post", V5, V4)) |>
select(-V5) |>
rename("V5" = "V51")
head(correctPost)## V2 V3 V4 V5
## 1 10 kiggif Post 13
## 2 10 kiggif Post 17
## 3 10 kiggif Post 26
## 4 10 kiggif AF2 28
## 5 10 kiggif AF2 52
## 6 10 kiggif MN1 1#rename new columns accordingly
nc<- correctPost %>%
rename("subj" = "V2",
"word" = "V3",
"block" = "V4",
"seq" = "V5")
head(nc)## subj word block seq
## 1 10 kiggif Post 13
## 2 10 kiggif Post 17
## 3 10 kiggif Post 26
## 4 10 kiggif AF2 28
## 5 10 kiggif AF2 52
## 6 10 kiggif MN1 1allcol<- as.data.frame(cbind(mfie$fileName, nc))|> rename("fileName"="mfie$fileName")
colnames(allcol)## [1] "fileName" "subj" "word" "block" "seq"df <- full_join(allcol, mfie, by = "fileName") |> rename("F0Hz" = "F0.Hz.", "F1Hz" = "F1.Hz.", "F2Hz" = "F2.Hz.", "F1Bark" = "F1.Bark.", "F2Bark" = "F2.Bark.")
#add column about BB condition
d <- df |>
mutate(
BB = case_when(
subj == "2" ~ "BB",
subj == "4" ~ "BB",
subj == "5" ~ "BB",
subj == "9" ~ "BB",
subj == "10" ~ "BB",
subj == "12" ~ "BB",
subj == "13" ~ "BB",
subj == "14" ~ "BB",
subj == "15" ~ "BB",
subj == "16" ~ "BB",
subj == "17" ~ "BB",
subj == "19" ~ "BB",
subj == "22" ~ "BB",
subj == "25" ~ "BB",
subj == "28" ~ "BB",
subj == "32" ~ "BB",
subj == "33" ~ "BB",
subj == "34" ~ "BB",
subj == "37" ~ "BB",
subj == "39" ~ "BB",
subj == "41" ~ "BB",
subj == "42" ~ "BB",
subj == "44" ~ "BB",
subj == "45" ~ "BB",
subj == "48" ~ "BB",
subj == "52" ~ "BB",
subj == "53" ~ "BB",
subj == "55" ~ "BB",
subj == "58" ~ "BB",
subj == "59" ~ "BB",
subj == "60" ~ "BB",
TRUE ~ "NB")
)plotting
library(ggplot2)
library(patchwork)
#NB group pre and post for I
nbprei<- d |> filter(BB == "NB", block == "Pre", Label == "I")
nbposti <- d |> filter(BB == "NB", block == "Post", Label == "I")
nbi <- rbind(nbprei, nbposti)
ggplot(data = nbi, aes(x = F2Hz, y = F1Hz, color = block))+
geom_point(alpha = 0.5) + xlim(1200, 3250) + ylim(100, 1750)
#NB group pre and post for E
nbpree<- d |> filter(BB == "NB", block == "Pre", Label == "E")
nbposte <- d |> filter(BB == "NB", block == "Post", Label == "E")
nbe <- rbind(nbpree, nbposte)
ggplot(data = nbe, aes(x = F2Hz, y = F1Hz, color = block))+
geom_point(alpha = 0.8) + xlim(1200, 3250) + ylim(100, 1750)
#I and E together
nbpre<- d |> filter(BB == "NB", block == "Pre")
nbpost <- d |> filter(BB == "NB", block == "Post")
nb <- rbind(nbpre, nbpost)
ggplot(data = nb, aes(x = F2Hz, y = F1Hz, color = block, shape = Label))+
geom_point(alpha = 0.7) +xlim(1200, 3250) + ylim(100, 1750)
pi<- ggplot(data = nbi, aes(x = F2Hz, y = F1Hz, color = block))+
geom_point(alpha = 0.5) +xlim(1200, 3250) + ylim(100, 1750) + labs(title = "pre vs post /I/")
pe<-ggplot(data = nbe, aes(x = F2Hz, y = F1Hz, color = block))+
geom_point(alpha = 0.8) +xlim(1200, 3250) + ylim(100, 1750)+ labs(title = "pre vs post /E/")
pi / pe
library(patchwork)
#BB group pre and post
bbpre<- d |> filter(BB == "BB", block == "Pre") #, Label == "E")
bbpost <- d |> filter(BB == "BB", block == "Post") #,Label == "E")
bb <- rbind(bbpre, bbpost)
#contrast at pretest for BB group
p1 <- ggplot(data = bbpre, aes(x = F2Hz, y = F1Hz, color = Label))+
geom_point(alpha = 0.5) + xlim(1200,3000) + ylim(100, 1500)+labs(title = "pre")
#contrast at posttest for BB group
p2 <- ggplot(data = bbpost, aes(x = F2Hz, y = F1Hz, color = Label))+
geom_point(alpha = 0.5) + xlim(1200,3000) + ylim(100, 1500) + labs(title = "post")
p1 + p2
counting subjects
d |> count(subj)## subj n
## 1 1 68
## 2 2 66
## 3 3 67
## 4 4 70
## 5 5 70
## 6 7 72
## 7 9 67
## 8 10 69
## 9 11 72
## 10 12 72
## 11 13 71
## 12 14 72
## 13 15 72
## 14 17 72
## 15 18 71
## 16 19 71
## 17 20 72
## 18 21 69
## 19 22 68
## 20 23 70
## 21 26 72
## 22 27 70
## 23 28 72
## 24 29 72
## 25 30 72
## 26 31 72
## 27 32 69
## 28 33 69
## 29 34 65
## 30 35 72
## 31 36 70
## 32 37 72
## 33 38 72
## 34 40 69
## 35 41 71
## 36 42 71
## 37 43 72
## 38 44 69
## 39 45 71
## 40 46 70
## 41 47 71
## 42 48 71
## 43 49 72
## 44 50 72
## 45 51 72
## 46 52 71
## 47 53 72
## 48 54 71
## 49 55 72
## 50 56 71
## 51 57 71
## 52 58 71
## 53 59 69
## 54 60 65coding in the block names correctly
E.g., if the participant went through the sequence of AF -> MN, then assign AF1, MN2 to the AF and MN blocks; if the participant went through the sequence of MN -> AF, then assign MN1, AF2 to the AF and MN blocks. Also coding in the bite block condition, since ultimately the analysis is comparing four groups.
#define a function
#correctBlocks <- function(n) {
mn2subj <- c("2", "3", "5", "6", "8", "9", "10", "12", "15", "17", "19", "26", "30", "31", "33", "35", "38", "40", "42", "43", "46", "48", "50", "51", "52", "53", "54", "56", "58", "60")
db <- d |>
mutate(
block = case_when(
subj %in% mn2subj & block == "MN1" ~ "MN2",
subj %in% mn2subj & block == "AF2" ~ "AF1",
subj %in% mn2subj & block == "Pre" ~ "Pre",
subj %in% mn2subj & block == "Post" ~ "Post",
!(subj %in% mn2subj) & block == "MN1" ~ "MN1",
!(subj %in% mn2subj) & block == "AF2" ~ "AF2",
!(subj %in% mn2subj) & block == "Pre" ~ "Pre",
!(subj %in% mn2subj) & block == "Post" ~ "Post"
)
)
#}#check if all subjects are coded correctly
db |> filter(block == "MN2") |> count(subj)## subj n
## 1 2 18
## 2 3 18
## 3 5 18
## 4 9 16
## 5 10 16
## 6 12 18
## 7 15 18
## 8 17 18
## 9 19 17
## 10 26 18
## 11 30 18
## 12 31 18
## 13 33 18
## 14 35 18
## 15 38 18
## 16 40 17
## 17 42 18
## 18 43 18
## 19 46 18
## 20 48 18
## 21 50 18
## 22 51 18
## 23 52 18
## 24 53 18
## 25 54 18
## 26 56 18
## 27 58 18
## 28 60 16library(ggplot2)
# /E/ in BB group pretest and test
bbprentest<- db |> filter(BB == "BB", Label == "E") |> filter(block == "Pre" | block == "AF1" | block == "MN1" | block == "Post")
#nrow(bbprentest)
ggplot(data = bbprentest, aes(x = F2Hz, y = F1Hz, color = block))+
geom_point(alpha = 0.5)
# /I/ in BB group pretest and test
bbprentest<- db |> filter(BB == "BB", Label == "I") |> filter(block == "Pre" | block == "AF1" | block == "MN1" | block == "Post")
#nrow(bbprentest)
ggplot(data = bbprentest, aes(x = F2Hz, y = F1Hz, color = block))+
geom_point(alpha = 0.5)
#BB group, pre vs post, both phonemes
library(patchwork)
bbprentestE<- db |> filter(BB == "BB", Label == "E") |> filter(block == "Pre" | block == "Post")
bbprentestI<- db |> filter(BB == "BB", Label == "I") |> filter(block == "Pre" | block == "Post")
#nrow(bbprentest)
p1 <- ggplot(data = bbprentestE, aes(x = F2Hz, y = F1Hz, color = block))+
geom_point(alpha = 0.5) + ggtitle("pre/post comparison of /E/")+theme(legend.position="none")
p2 <- ggplot(data = bbprentestI, aes(x = F2Hz, y = F1Hz, color = block))+
geom_point(alpha = 0.5)+ ggtitle("pre/post comparison of /I/")
p1 + p2 
linear regression (all blocks):
#blocks: pre, AF1, MN1, Post
library(lme4)## Loading required package: Matrixaf1subj <- c("2", "3", "5", "6", "8", "9", "10", "12", "15", "17", "19", "26", "30", "31", "33", "35", "38", "40", "42", "43", "46", "48", "50", "51", "52", "53", "54", "56", "58", "60")
df <- db |> filter(block == "Pre" | block == "AF1" | block == "MN1" | block == "Post") |>
mutate(
AF = case_when(
subj %in% af1subj ~ "AF",
TRUE ~ "MN")) |> mutate(
block = case_when(
block == "AF1" ~ "2",
block == "MN1" ~ "2",
block == "Pre" ~ "1",
block == "Post" ~ "4")
)
#change variables into correct data types
df$BB <- as.factor(df$BB)
df$Label <- as.factor(df$Label)
df$block <- as.factor(df$block)
df$gender <- as.factor(df$gender)
df$AF <- as.factor(df$AF)
df$subj <- as.character(df$subj)
df$seq <- as.numeric(df$seq)
df$F0Hz <- as.numeric(df$F0Hz)## Warning: NAs introduced by coerciondf$F1Hz <- as.numeric(df$F1Hz)
df$F2Hz <- as.numeric(df$F2Hz)
df$F1Bark <- as.numeric(df$F1Bark)
df$F2Bark <- as.numeric(df$F2Bark)
write.csv(df, "productionFullTable.csv")
#contrast coding
df$BB <- relevel(df$BB, ref = "NB")
df$Label <- relevel(df$Label, ref = "I")
df$block <- relevel(df$block, ref = "1")
df$gender <- relevel(df$gender, ref = "F")
#model with blocks 1,2,4
f1hzallbl <- lmer(F1Hz ~ block * BB * AF* Label + gender + (1|subj) + (1|word), data = df)
summary(f1hzallbl)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Hz ~ block * BB * AF * Label + gender + (1 | subj) + (1 | word)
## Data: df
##
## REML criterion at convergence: 32608.2
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.4079 -0.5143 0.0383 0.5118 16.4201
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 3096.4 55.65
## word (Intercept) 132.3 11.50
## Residual 5493.9 74.12
## Number of obs: 2847, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 450.7640 18.8082 23.966
## block2 0.6292 9.7768 0.064
## block4 14.8829 9.7121 1.532
## BBBB 9.2792 23.1196 0.401
## AFMN -18.6968 24.1381 -0.775
## LabelE 147.4586 13.5919 10.849
## genderM -96.9882 21.2561 -4.563
## block2:BBBB 36.7918 13.3972 2.746
## block4:BBBB -24.5214 13.3629 -1.835
## block2:AFMN 63.0118 13.8559 4.548
## block4:AFMN -28.1106 13.7801 -2.040
## BBBB:AFMN -9.7971 33.2636 -0.295
## block2:LabelE -8.1994 13.8630 -0.591
## block4:LabelE -18.4500 13.8138 -1.336
## BBBB:LabelE -20.8307 13.4341 -1.551
## AFMN:LabelE 39.3578 13.8591 2.840
## block2:BBBB:AFMN -23.3488 19.3244 -1.208
## block4:BBBB:AFMN 36.4373 19.2392 1.894
## block2:BBBB:LabelE -9.4693 18.9657 -0.499
## block4:BBBB:LabelE 30.9208 18.9483 1.632
## block2:AFMN:LabelE 35.7435 19.5786 1.826
## block4:AFMN:LabelE 27.4315 19.5549 1.403
## BBBB:AFMN:LabelE -23.8587 19.2555 -1.239
## block2:BBBB:AFMN:LabelE -29.3712 27.2894 -1.076
## block4:BBBB:AFMN:LabelE -36.1044 27.2431 -1.325##
## Correlation matrix not shown by default, as p = 25 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need itInterpretations:
Estimate Std. Error t value
block2 0.6292 9.7768 0.064 in the NB group, F1 in /I/ at test did not differ from pretest, for female speakers
block4 14.8829 9.7121 1.532 in the NB group, F1 in /I/ at posttest did not differ from pretest, for female speakers
BBBB 9.2792 23.1196 0.401 the F1 in /I/ in BB group did not differ from the NB group at pretest, for female speakers
AFMN -18.6968 24.1381 -0.775 the F1 in /I/ in the MN group did not differ from the AF group at pretest, for female speakers
LabelE 147.4586 13.5919 10.849 the F1 for /E/ in the NB group is 147.4586 Hz higher than /I/ at pretest, for female speakers
genderM -96.9882 21.2561 -4.563 at pretest, male speakers have an F1 in /I/ that is 96.9882 Hz lower than that of female speakers
block2:BBBB 36.7918 13.3972 2.746 the F1 increase in /I/ from pretest to test is 36.7918 Hz higher in the BB group than in the NB and AF group, for female speakers
block4:BBBB -24.5214 13.3629 -1.835 the difference in F1 from pretest to posttest did not differ between BB and NB group (female speakers)
block2:AFMN 63.0118 13.8559 4.548 from pretest to test, the F1 in /I/ in the NBMN group increased 63.0118 Hz more than the NBAF group (f speakers)
block4:AFMN -28.1106 13.7801 -2.040 at posttest the F1 in /I/ in the MN group is 28.11 Hz lower than at pretest
BBBB:AFMN -9.7971 33.2636 -0.295 the F1 in /I/ did not differ between BBAF and BBMN group at pretest
block2:LabelE -8.1994 13.8630 -0.591 the F1 difference between /I/ and /E/ is not bigger in the NB group when comparing test to pretest
block4:LabelE -18.4500 13.8138 -1.336 the difference in F1 between /I/ and /E/ did not change from pre to posttest for the NB group
BBBB:LabelE -20.8307 13.4341 -1.551 the difference in F1 between /I/ and /E/ did not differ between BB and NB group at pretest
AFMN:LabelE 39.3578 13.8591 2.840 the difference in F1 between /I/ and /E/ is 39 Hz bigger in the NBMN group than in the NBAF group at pretest
block2:BBBB:AFMN -23.3488 19.3244 -1.208 when comparing test to pretest, the AF/MN did not affect the effect that BB has on the F1 of /I/
block4:BBBB:AFMN 36.4373 19.2392 1.894 when comparing posttest to pretest, the AF/MN did not affect the effect that BB has on the F1 of /I/
block2:BBBB:LabelE -9.4693 18.9657 -0.499 when comparing test to pretest, BB did not affect the difference between /I/ and /E/ in the AF group
block4:BBBB:LabelE 30.9208 18.9483 1.632 when comparing posttest to pretest, BB did not affect the difference between /I/ and /E/ in the AF group
block2:AFMN:LabelE 35.7435 19.5786 1.826 when comparing test to pretest, MN did not affect the difference between /I/ and /E/ in the NB group
block4:AFMN:LabelE 27.4315 19.5549 1.403 when comparing posttest to pretest, MN did not affect the difference between /I/ and /E/ in the NB group
BBBB:AFMN:LabelE -23.8587 19.2555 -1.239 at pretest, the difference between /I/ and /E/ caused by being in the BB group (not significant) was not modulated by being assigned in the MN group
block2:BBBB:AFMN:LabelE -29.3712 27.2894 -1.076 when comparing test to pretest, the effect of BB on the /I/-/E/ contrast (not significant, see block2:BBBB:LabelE) was not modulated by the MN
block4:BBBB:AFMN:LabelE -36.1044 27.2431 -1.325 when comparing posttest to pretest, the effect of BB on the /I/-/E/ contrast (not significant, see block4:BBBB:LabelE) was not modulated by the MN
#effect plots
library (effects)## Loading required package: carData## lattice theme set by effectsTheme()
## See ?effectsTheme for details. eall.f1hzallbl <- predictorEffects(f1hzallbl)
#plot(eall.f1hzallbl)
#plot(predictorEffects(f1hzallbl, ~BB))
#plot(predictorEffects(f1hzallbl, ~block))
plot(predictorEffects(f1hzallbl, ~Label))
# BB group, pre vs test, both phonemes
bbprentestE<- df |> filter(BB == "BB", Label == "E") |> filter(block == "1" | block == "2") |> mutate(block = case_when(block == "1" ~ "pretest",
block == "2" ~ "test"))
bbprentestI<- df |> filter(BB == "BB", Label == "I") |> filter(block == "1" | block == "2")|> mutate(block = case_when(block == "1" ~ "pretest",
block == "2" ~ "test"))
#nrow(bbprentest)
p1 <- ggplot(data = bbprentestE, aes(x = F2Hz, y = F1Hz, color = block))+
geom_point(alpha = 0.5) + ylim (200, 1500) + ggtitle("bite block effect for /E/")+theme(legend.position="none")
p2 <- ggplot(data = bbprentestI, aes(x = F2Hz, y = F1Hz, color = block))+
geom_point(alpha = 0.5) + ylim (200, 1500) + ggtitle("bite block effect for /I/")
p1 + p2 ## Warning: Removed 1 row containing missing values or values outside the scale range
## (`geom_point()`).
#### similar model on F2
f2hzallbl <- lmer(F2Hz ~ block * BB * AF* Label + gender + (1|subj) + (1|word), data = df)
summary(f2hzallbl)## Linear mixed model fit by REML ['lmerMod']
## Formula: F2Hz ~ block * BB * AF * Label + gender + (1 | subj) + (1 | word)
## Data: df
##
## REML criterion at convergence: 35901.7
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.2993 -0.5454 -0.0166 0.5149 5.8049
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 11216 105.90
## word (Intercept) 9938 99.69
## Residual 17535 132.42
## Number of obs: 2847, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 2200.053 66.434 33.116
## block2 -9.267 17.467 -0.531
## block4 -56.774 17.351 -3.272
## BBBB -79.686 43.562 -1.829
## AFMN 5.520 45.488 0.121
## LabelE -255.947 83.267 -3.074
## genderM -239.801 40.375 -5.939
## block2:BBBB -651.743 23.935 -27.230
## block4:BBBB 28.908 23.874 1.211
## block2:AFMN -54.598 24.754 -2.206
## block4:AFMN 46.824 24.619 1.902
## BBBB:AFMN 97.834 62.676 1.561
## block2:LabelE -15.246 24.767 -0.616
## block4:LabelE 43.164 24.679 1.749
## BBBB:LabelE 65.757 24.001 2.740
## AFMN:LabelE 6.270 24.760 0.253
## block2:BBBB:AFMN -85.069 34.524 -2.464
## block4:BBBB:AFMN -60.424 34.372 -1.758
## block2:BBBB:LabelE 190.195 33.883 5.613
## block4:BBBB:LabelE -38.486 33.852 -1.137
## block2:AFMN:LabelE 36.238 34.978 1.036
## block4:AFMN:LabelE -33.284 34.936 -0.953
## BBBB:AFMN:LabelE -62.233 34.401 -1.809
## block2:BBBB:AFMN:LabelE 8.991 48.754 0.184
## block4:BBBB:AFMN:LabelE 25.526 48.672 0.524##
## Correlation matrix not shown by default, as p = 25 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need itSignificant results:
Estimate Std. Error t value
block4 -56.774 17.351 -3.272 in the NB group, F2 in /I/ at posttest is 56.774 Hz lower than pretest, for female speakers
LabelE -255.947 83.267 -3.074 at pretest female speakers’ F2 in /E/ is 255.947 Hz lower than their F2 in /I/
genderM -239.801 40.375 -5.939 at pretest female speakers’ F2 in /I/ 239.801 Hz lower than the females speakers
block2:BBBB -651.743 23.935 -27.230 the change of F2 in /I/ from pretest to test is 651.743 Hz smaller in the BB group when compared to BB group (which did not change) ==> the F2 in BB group decreased 651.743 Hz at test when compared to pretest (for female speakers)
block4:BBBB -24.5214 13.3629 -1.835 the difference in F1 from pretest to posttest did not differ between BB and NB group (female speakers)
block4:BBBB 28.908 23.874 1.211 the change in F2 in /I/ from pretest to posttest (sig. in NB group) is not modulated by being in the BB group (having BB did not change the “base” effect) (for female speakers)
block2:AFMN -54.598 24.754 -2.206 the change in F2 in /I/ for the NB group from pretest to test (not sig.) isn ot modualted by MN (for female speakers)
block4:AFMN 46.824 24.619 1.902 the change in F2 in /I/ for the NB from pretest to posttest (sig.) is not modulated by MN (having MN did not change the “base” effect) (for female speakers)
BBBB:LabelE 65.757 24.001 2.740 the F2 difference between /I/ and /E/ is 65.757 Hz higher in the BB group than the NB group at pretest (for female speakers)
block2:BBBB:AFMN -85.069 34.524 -2.464 having MN decreased the F2 shift from test to pretest in the BB group (sig.) for 85.069 Hz (for female speakers)
block2:BBBB:LabelE 190.195 33.883 5.613 when comparing test to pretest, BB increased the F2 difference between /I/ and /E/ in the AF group (sig.) for 190.195 Hz (for female speakers)
block4:BBBB:LabelE -38.486 33.852 -1.137 when comparing posttest to pretest, having had the BB did not affect the F2 difference between /I/ and /E/ in the AF group (not sig.)(for female speakers).
block2:AFMN:LabelE 36.238 34.978 1.036 MN did not affect the /I/-/E/ contrast in F2 in the NB group at test (for female speakers)
block4:AFMN:LabelE -33.284 34.936 -0.953 MN did not affect the /I/-/E/ contrast in F2 in the NB group at posttest (for female speakers)
block2:BBBB:AFMN:LabelE 8.991 48.754 0.184 MN did not affect the effect that BB has on the /I/-/E/ contrast in F2 at test (for female speakers) block4:BBBB:AFMN:LabelE 25.526 48.672 0.524 MN did not affect the effect that BB has on the /I/-/E/ contrast in F2 at posttest (for female speakers)
model including sequence number as a variable
#convert seq as numeric instead of categorical
df$seq <- as.numeric(df$seq)#seq as a separate variable
f1hzallblsq1 <- lmer(F1Hz ~ block * BB * AF* Label + seq + gender + (1|subj) + (1|word), data = df)
summary(f1hzallblsq1)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Hz ~ block * BB * AF * Label + seq + gender + (1 | subj) +
## (1 | word)
## Data: df
##
## REML criterion at convergence: 32609.4
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.4209 -0.5126 0.0397 0.5079 16.4523
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 3096 55.64
## word (Intercept) 134 11.58
## Residual 5492 74.11
## Number of obs: 2847, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 453.94339 18.97583 23.922
## block2 0.68420 9.77555 0.070
## block4 14.81309 9.71091 1.525
## BBBB 8.98581 23.11781 0.389
## AFMN -18.81789 24.13532 -0.780
## LabelE 147.18887 13.63335 10.796
## seq -0.11784 0.08967 -1.314
## genderM -97.08227 21.25369 -4.568
## block2:BBBB 37.01150 13.39637 2.763
## block4:BBBB -24.28797 13.36223 -1.818
## block2:AFMN 63.05772 13.85402 4.552
## block4:AFMN -27.84721 13.77966 -2.021
## BBBB:AFMN -9.28651 33.26181 -0.279
## block2:LabelE -8.27694 13.86126 -0.597
## block4:LabelE -18.10164 13.81444 -1.310
## BBBB:LabelE -20.14902 13.44223 -1.499
## AFMN:LabelE 39.79007 13.86106 2.871
## block2:BBBB:AFMN -23.86539 19.32573 -1.235
## block4:BBBB:AFMN 35.97379 19.23983 1.870
## block2:BBBB:LabelE -9.95622 18.96677 -0.525
## block4:BBBB:LabelE 30.28566 18.95189 1.598
## block2:AFMN:LabelE 35.85172 19.57609 1.831
## block4:AFMN:LabelE 26.78403 19.55841 1.369
## BBBB:AFMN:LabelE -24.63411 19.26195 -1.279
## block2:BBBB:AFMN:LabelE -29.13322 27.28627 -1.068
## block4:BBBB:AFMN:LabelE -35.40983 27.24453 -1.300##
## Correlation matrix not shown by default, as p = 26 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need itseq itself is not significant.
#seq as interaction in hertz model
f1hzallblsq <- lmer(F1Hz ~ block * BB * AF* Label * seq + gender + (1|subj) + (1|word), data = df)
summary(f1hzallblsq)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Hz ~ block * BB * AF * Label * seq + gender + (1 | subj) +
## (1 | word)
## Data: df
##
## REML criterion at convergence: 32561.2
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.5479 -0.5204 0.0382 0.5162 16.2347
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 3097 55.65
## word (Intercept) 137 11.70
## Residual 5458 73.88
## Number of obs: 2847, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 453.569002 22.383403 20.264
## block2 0.694851 19.918535 0.035
## block4 -12.465988 19.538823 -0.638
## BBBB 12.308602 27.918045 0.441
## AFMN -14.629281 29.235276 -0.500
## LabelE 127.285342 21.333875 5.966
## seq -0.105443 0.449356 -0.235
## genderM -96.853478 21.256326 -4.556
## block2:BBBB 21.667816 26.374745 0.822
## block4:BBBB 1.384655 25.922118 0.053
## block2:AFMN 64.403717 27.563602 2.337
## block4:AFMN -13.699684 27.430525 -0.499
## BBBB:AFMN -24.363365 40.468682 -0.602
## block2:LabelE 9.727838 27.361416 0.356
## block4:LabelE 35.254971 27.218245 1.295
## BBBB:LabelE 0.532953 26.262855 0.020
## AFMN:LabelE 47.316420 26.749871 1.769
## block2:seq -0.001075 0.641695 -0.002
## block4:seq 1.040861 0.640452 1.625
## BBBB:seq -0.136151 0.610426 -0.223
## AFMN:seq -0.160133 0.625337 -0.256
## LabelE:seq 0.810170 0.635859 1.274
## block2:BBBB:AFMN 10.165262 38.112746 0.267
## block4:BBBB:AFMN -0.794306 38.286573 -0.021
## block2:BBBB:LabelE -5.322872 37.090110 -0.144
## block4:BBBB:LabelE -13.339072 37.342183 -0.357
## block2:AFMN:LabelE 23.163589 39.017167 0.594
## block4:AFMN:LabelE -15.719360 38.316997 -0.410
## BBBB:AFMN:LabelE -1.756773 37.952125 -0.046
## block2:BBBB:seq 0.588238 0.863211 0.681
## block4:BBBB:seq -0.971644 0.861193 -1.128
## block2:AFMN:seq -0.045073 0.892541 -0.050
## block4:AFMN:seq -0.555470 0.891860 -0.623
## BBBB:AFMN:seq 0.569734 0.877349 0.649
## block2:LabelE:seq -0.730490 0.915852 -0.798
## block4:LabelE:seq -2.081253 0.898202 -2.317
## BBBB:LabelE:seq -0.823171 0.870897 -0.945
## AFMN:LabelE:seq -0.348976 0.876285 -0.398
## block2:BBBB:AFMN:LabelE -37.480605 54.123379 -0.693
## block4:BBBB:AFMN:LabelE 2.850314 54.108715 0.053
## block2:BBBB:AFMN:seq -1.295221 1.240900 -1.044
## block4:BBBB:AFMN:seq 1.370336 1.247733 1.098
## block2:BBBB:LabelE:seq -0.158216 1.231770 -0.128
## block4:BBBB:LabelE:seq 1.703560 1.230058 1.385
## block2:AFMN:LabelE:seq 0.529280 1.276857 0.415
## block4:AFMN:LabelE:seq 1.693677 1.247495 1.358
## BBBB:AFMN:LabelE:seq -0.744507 1.228994 -0.606
## block2:BBBB:AFMN:LabelE:seq 0.118929 1.770825 0.067
## block4:BBBB:AFMN:LabelE:seq -1.555320 1.756269 -0.886##
## Correlation matrix not shown by default, as p = 49 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need itseq itself is not significant. The only significant
interaction is block4:LabelE:seq. Interpretation: the
difference of the /I/-/E/ contrast between block 2 and block 4 (diff of
diff) is modified by the sequence. As sequence increase 1, the contrast
decreases by 2.85 Hz. But since the intercept is on the NB group. This
is only about the group without bite blocks.
block4:BBBB:LabelE:seq is not significant.
#try releveling the reference level to BB to check if block4:LabelE:seq is significant
df$BB <- relevel(df$BB, ref = "BB")
f1hzallblsqrl <- lmer(F1Hz ~ block * BB * AF* Label * seq + gender + (1|subj) + (1|word), data = df)
summary(f1hzallblsqrl)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Hz ~ block * BB * AF * Label * seq + gender + (1 | subj) +
## (1 | word)
## Data: df
##
## REML criterion at convergence: 32561.2
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.5479 -0.5204 0.0382 0.5162 16.2347
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 3097 55.65
## word (Intercept) 137 11.70
## Residual 5458 73.88
## Number of obs: 2847, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 465.87760 20.64867 22.562
## block2 22.36267 17.29093 1.293
## block4 -11.08133 17.03638 -0.650
## BBNB -12.30860 27.91804 -0.441
## AFMN -38.99265 28.47355 -1.369
## LabelE 127.81830 20.42798 6.257
## seq -0.24159 0.41331 -0.585
## genderM -96.85348 21.25633 -4.556
## block2:BBNB -21.66782 26.37475 -0.822
## block4:BBNB -1.38466 25.92212 -0.053
## block2:AFMN 74.56898 26.33223 2.832
## block4:AFMN -14.49399 26.70058 -0.543
## BBNB:AFMN 24.36336 40.46868 0.602
## block2:LabelE 4.40497 25.04228 0.176
## block4:LabelE 21.91590 25.54744 0.858
## BBNB:LabelE -0.53295 26.26285 -0.020
## AFMN:LabelE 45.55965 26.93729 1.691
## block2:seq 0.58716 0.57742 1.017
## block4:seq 0.06922 0.57574 0.120
## BBNB:seq 0.13615 0.61043 0.223
## AFMN:seq 0.40960 0.61600 0.665
## LabelE:seq -0.01300 0.59514 -0.022
## block2:BBNB:AFMN -10.16526 38.11275 -0.267
## block4:BBNB:AFMN 0.79431 38.28657 0.021
## block2:BBNB:LabelE 5.32287 37.09011 0.144
## block4:BBNB:LabelE 13.33907 37.34218 0.357
## block2:AFMN:LabelE -14.31702 37.51885 -0.382
## block4:AFMN:LabelE -12.86905 38.18543 -0.337
## BBNB:AFMN:LabelE 1.75677 37.95212 0.046
## block2:BBNB:seq -0.58824 0.86321 -0.681
## block4:BBNB:seq 0.97164 0.86119 1.128
## block2:AFMN:seq -1.34029 0.86258 -1.554
## block4:AFMN:seq 0.81487 0.87235 0.934
## BBNB:AFMN:seq -0.56973 0.87735 -0.649
## block2:LabelE:seq -0.88871 0.82372 -1.079
## block4:LabelE:seq -0.37769 0.83956 -0.450
## BBNB:LabelE:seq 0.82317 0.87090 0.945
## AFMN:LabelE:seq -1.09348 0.86245 -1.268
## block2:BBNB:AFMN:LabelE 37.48060 54.12338 0.693
## block4:BBNB:AFMN:LabelE -2.85031 54.10871 -0.053
## block2:BBNB:AFMN:seq 1.29522 1.24090 1.044
## block4:BBNB:AFMN:seq -1.37034 1.24773 -1.098
## block2:BBNB:LabelE:seq 0.15822 1.23177 0.128
## block4:BBNB:LabelE:seq -1.70356 1.23006 -1.385
## block2:AFMN:LabelE:seq 0.64821 1.22734 0.528
## block4:AFMN:LabelE:seq 0.13836 1.23554 0.112
## BBNB:AFMN:LabelE:seq 0.74451 1.22899 0.606
## block2:BBNB:AFMN:LabelE:seq -0.11893 1.77082 -0.067
## block4:BBNB:AFMN:LabelE:seq 1.55532 1.75627 0.886##
## Correlation matrix not shown by default, as p = 49 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need itNot.
#bark model with seq
#correct back to NB as reference
df$BB <- relevel(df$BB, ref = "NB")
f1bkallblsq <- lmer(F1Bark ~ block * BB * AF* Label * seq + gender + (1|subj) + (1|word), data = df)
summary(f1bkallblsq)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Bark ~ block * BB * AF * Label * seq + gender + (1 | subj) +
## (1 | word)
## Data: df
##
## REML criterion at convergence: 6071.3
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.2400 -0.6094 0.0031 0.5904 11.5905
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 0.19863 0.44568
## word (Intercept) 0.00688 0.08294
## Residual 0.42361 0.65086
## Number of obs: 2847, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 4.5809941 0.1846953 24.803
## block2 -0.0422332 0.1754793 -0.241
## block4 -0.0760253 0.1721333 -0.442
## BBBB 0.0433653 0.2334445 0.186
## AFMN -0.1460593 0.2444995 -0.597
## LabelE 1.0536870 0.1811664 5.816
## seq -0.0029714 0.0039587 -0.751
## genderM -0.7973375 0.1708296 -4.667
## block2:BBBB 0.2289985 0.2323565 0.986
## block4:BBBB -0.0507522 0.2283634 -0.222
## block2:AFMN 0.5748125 0.2428270 2.367
## block4:AFMN -0.2152174 0.2416602 -0.891
## BBBB:AFMN -0.1059761 0.3386624 -0.313
## block2:LabelE 0.0868794 0.2410493 0.360
## block4:LabelE 0.2318649 0.2397802 0.967
## BBBB:LabelE 0.0123174 0.2313694 0.053
## AFMN:LabelE 0.4032193 0.2356614 1.711
## block2:seq 0.0008324 0.0056532 0.147
## block4:seq 0.0057264 0.0056422 1.015
## BBBB:seq 0.0015958 0.0053777 0.297
## AFMN:seq -0.0010558 0.0055091 -0.192
## LabelE:seq 0.0069009 0.0056016 1.232
## block2:BBBB:AFMN -0.0598676 0.3357556 -0.178
## block4:BBBB:AFMN 0.0711472 0.3372809 0.211
## block2:BBBB:LabelE -0.0391282 0.3267573 -0.120
## block4:BBBB:LabelE 0.0252423 0.3289558 0.077
## block2:AFMN:LabelE 0.0967545 0.3437245 0.281
## block4:AFMN:LabelE 0.0318042 0.3375588 0.094
## BBBB:AFMN:LabelE -0.0410696 0.3343484 -0.123
## block2:BBBB:seq 0.0032898 0.0076046 0.433
## block4:BBBB:seq -0.0073793 0.0075866 -0.973
## block2:AFMN:seq 0.0019215 0.0078629 0.244
## block4:AFMN:seq -0.0005979 0.0078571 -0.076
## BBBB:AFMN:seq 0.0040791 0.0077290 0.528
## block2:LabelE:seq -0.0052679 0.0080684 -0.653
## block4:LabelE:seq -0.0127973 0.0079126 -1.617
## BBBB:LabelE:seq -0.0076302 0.0076723 -0.995
## AFMN:LabelE:seq -0.0050471 0.0077198 -0.654
## block2:BBBB:AFMN:LabelE -0.1593360 0.4768049 -0.334
## block4:BBBB:AFMN:LabelE -0.2010132 0.4766596 -0.422
## block2:BBBB:AFMN:seq -0.0123218 0.0109315 -1.127
## block4:BBBB:AFMN:seq 0.0101097 0.0109915 0.920
## block2:BBBB:LabelE:seq -0.0021582 0.0108515 -0.199
## block4:BBBB:LabelE:seq 0.0115886 0.0108356 1.069
## block2:AFMN:LabelE:seq 0.0016732 0.0112483 0.149
## block4:AFMN:LabelE:seq 0.0113822 0.0109897 1.036
## BBBB:AFMN:LabelE:seq -0.0042692 0.0108270 -0.394
## block2:BBBB:AFMN:LabelE:seq 0.0031307 0.0156000 0.201
## block4:BBBB:AFMN:LabelE:seq -0.0112874 0.0154711 -0.730##
## Correlation matrix not shown by default, as p = 49 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need itNothing involving seq is significant in the bark
model.
#F2 model in bark
f2hzallblsq <- lmer(F2Hz ~ block * BB * AF* Label * seq + gender + (1|subj) + (1|word), data = df)
summary(f2hzallblsq)## Linear mixed model fit by REML ['lmerMod']
## Formula: F2Hz ~ block * BB * AF * Label * seq + gender + (1 | subj) +
## (1 | word)
## Data: df
##
## REML criterion at convergence: 35824.5
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.3893 -0.5403 -0.0210 0.5273 5.6621
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 11310 106.35
## word (Intercept) 9826 99.13
## Residual 17403 131.92
## Number of obs: 2847, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 2198.63474 69.63184 31.575
## block2 7.82478 35.57051 0.220
## block4 -86.34025 34.89222 -2.474
## BBBB -88.78280 51.88319 -1.711
## AFMN -35.76408 54.32488 -0.658
## LabelE -232.64020 87.81167 -2.649
## seq 0.05726 0.80246 0.071
## genderM -240.30049 40.53803 -5.928
## block2:BBBB -645.83112 47.10080 -13.712
## block4:BBBB 76.74581 46.29401 1.658
## block2:AFMN -36.15632 49.22528 -0.735
## block4:AFMN 133.29846 48.98492 2.721
## BBBB:AFMN 138.80221 75.16615 1.847
## block2:LabelE -46.25654 48.86237 -0.947
## block4:LabelE 61.55774 48.60761 1.266
## BBBB:LabelE 56.27668 46.90091 1.200
## AFMN:LabelE 0.49400 47.77022 0.010
## block2:seq -0.62846 1.14595 -0.548
## block4:seq 1.12952 1.14372 0.988
## BBBB:seq 0.38247 1.09012 0.351
## AFMN:seq 1.59333 1.11674 1.427
## LabelE:seq -0.94468 1.13556 -0.832
## block2:BBBB:AFMN -76.02562 68.06706 -1.117
## block4:BBBB:AFMN -116.37837 68.37955 -1.702
## block2:BBBB:LabelE 202.06710 66.23631 3.051
## block4:BBBB:LabelE -30.21319 66.69212 -0.453
## block2:AFMN:LabelE 34.77917 69.67939 0.499
## block4:AFMN:LabelE -62.84834 68.42744 -0.918
## BBBB:AFMN:LabelE -55.17415 67.77672 -0.814
## block2:BBBB:seq -0.27548 1.54158 -0.179
## block4:BBBB:seq -1.86635 1.53805 -1.213
## block2:AFMN:seq -0.75118 1.59401 -0.471
## block4:AFMN:seq -3.29466 1.59268 -2.069
## BBBB:AFMN:seq -1.64068 1.56688 -1.047
## block2:LabelE:seq 1.19317 1.63556 0.730
## block4:LabelE:seq -0.63627 1.60407 -0.397
## BBBB:LabelE:seq 0.38826 1.55529 0.250
## AFMN:LabelE:seq 0.21977 1.56491 0.140
## block2:BBBB:AFMN:LabelE 67.28206 96.65795 0.696
## block4:BBBB:AFMN:LabelE -32.93399 96.63682 -0.341
## block2:BBBB:AFMN:seq -0.26172 2.21623 -0.118
## block4:BBBB:AFMN:seq 2.25621 2.22851 1.012
## block2:BBBB:LabelE:seq -0.45108 2.19976 -0.205
## block4:BBBB:LabelE:seq -0.30785 2.19694 -0.140
## block2:AFMN:LabelE:seq 0.04045 2.28034 0.018
## block4:AFMN:LabelE:seq 1.09435 2.22785 0.491
## BBBB:AFMN:LabelE:seq -0.20467 2.19483 -0.093
## block2:BBBB:AFMN:LabelE:seq -2.44684 3.16254 -0.774
## block4:BBBB:AFMN:LabelE:seq 2.09309 3.13675 0.667##
## Correlation matrix not shown by default, as p = 49 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need itNothing involving seq is significant.
#F2 model in bark
f2bkallblsq <- lmer(F2Bark ~ block * BB * AF* Label * seq + gender + (1|subj) + (1|word), data = df)
summary(f2bkallblsq)## Linear mixed model fit by REML ['lmerMod']
## Formula: F2Bark ~ block * BB * AF * Label * seq + gender + (1 | subj) +
## (1 | word)
## Data: df
##
## REML criterion at convergence: 5385.3
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.3755 -0.5544 0.0030 0.6030 3.6936
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 0.13158 0.3627
## word (Intercept) 0.09797 0.3130
## Residual 0.33111 0.5754
## Number of obs: 2847, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 13.5804002 0.2356460 57.631
## block2 -0.1104026 0.1551391 -0.712
## block4 -0.2654031 0.1521764 -1.744
## BBBB -0.3211376 0.1979699 -1.622
## AFMN -0.1073032 0.2073753 -0.517
## LabelE -0.7667017 0.2956099 -2.594
## seq -0.0018356 0.0034997 -0.525
## genderM -0.8816616 0.1395343 -6.319
## block2:BBBB -2.3915642 0.2054309 -11.642
## block4:BBBB 0.2583066 0.2019062 1.279
## block2:AFMN -0.0720172 0.2146976 -0.335
## block4:AFMN 0.3278160 0.2136461 1.534
## BBBB:AFMN 0.3399638 0.2874145 1.183
## block2:LabelE 0.0184801 0.2131119 0.087
## block4:LabelE 0.1654170 0.2119821 0.780
## BBBB:LabelE 0.1734905 0.2045541 0.848
## AFMN:LabelE -0.0656063 0.2083470 -0.315
## block2:seq 0.0026565 0.0049979 0.532
## block4:seq 0.0037764 0.0049880 0.757
## BBBB:seq 0.0033084 0.0047544 0.696
## AFMN:seq 0.0049168 0.0048705 1.010
## LabelE:seq 0.0001313 0.0049525 0.027
## block2:BBBB:AFMN -0.0619287 0.2968747 -0.209
## block4:BBBB:AFMN -0.2845105 0.2982344 -0.954
## block2:BBBB:LabelE 0.3876647 0.2888899 1.342
## block4:BBBB:LabelE -0.0666004 0.2908500 -0.229
## block2:AFMN:LabelE -0.0402135 0.3038880 -0.132
## block4:AFMN:LabelE -0.1721993 0.2984237 -0.577
## BBBB:AFMN:LabelE -0.0175606 0.2956062 -0.059
## block2:BBBB:seq -0.0066366 0.0067235 -0.987
## block4:BBBB:seq -0.0075953 0.0067078 -1.132
## block2:AFMN:seq -0.0048580 0.0069522 -0.699
## block4:AFMN:seq -0.0089318 0.0069462 -1.286
## BBBB:AFMN:seq -0.0014945 0.0068338 -0.219
## block2:LabelE:seq -0.0044310 0.0071334 -0.621
## block4:LabelE:seq -0.0043017 0.0069952 -0.615
## BBBB:LabelE:seq -0.0032561 0.0067832 -0.480
## AFMN:LabelE:seq -0.0002505 0.0068251 -0.037
## block2:BBBB:AFMN:LabelE 0.2719617 0.4215589 0.645
## block4:BBBB:AFMN:LabelE -0.1559470 0.4214541 -0.370
## block2:BBBB:AFMN:seq -0.0056539 0.0096659 -0.585
## block4:BBBB:AFMN:seq 0.0050381 0.0097192 0.518
## block2:BBBB:LabelE:seq 0.0133406 0.0095941 1.391
## block4:BBBB:LabelE:seq 0.0036973 0.0095807 0.386
## block2:AFMN:LabelE:seq 0.0082206 0.0099448 0.827
## block4:AFMN:LabelE:seq 0.0078683 0.0097155 0.810
## BBBB:AFMN:LabelE:seq -0.0028381 0.0095725 -0.296
## block2:BBBB:AFMN:LabelE:seq -0.0162579 0.0137926 -1.179
## block4:BBBB:AFMN:LabelE:seq 0.0030418 0.0136795 0.222##
## Correlation matrix not shown by default, as p = 49 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need itNothing involving seq is significant.
interaction plots
#model for pretest vs test in order to only plot the interaction effect in pre vs post
prentest <- df |> filter(block == "1" | block == "2")|> mutate(block = case_when(block == "1" ~ "pretest",
block == "2" ~ "test"))
f2hzprentest <- lmer(F2Hz ~ block * BB * AF* Label + gender + (1|subj) + (1|word), data = prentest)
#effect plot
library(effects)
plot(effect( "block * BB * AF", f2hzprentest))## NOTE: block:BB:AF is not a high-order term in the model
#bark model
f1barkallbl <- lmer(F1Bark ~ block * BB * AF* Label + gender + (1|subj) + (1|word), data = df)
summary(f1barkallbl)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Bark ~ block * BB * AF * Label + gender + (1 | subj) + (1 |
## word)
## Data: df
##
## REML criterion at convergence: 5888.2
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.1009 -0.6093 0.0061 0.6128 11.6410
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 0.199013 0.44611
## word (Intercept) 0.006484 0.08052
## Residual 0.425990 0.65268
## Number of obs: 2847, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 4.501109 0.150669 29.874
## block2 -0.020860 0.086091 -0.242
## block4 0.075867 0.085521 0.887
## BBBB 0.089611 0.188519 0.475
## AFMN -0.170527 0.196775 -0.867
## LabelE 1.230121 0.108655 11.321
## genderM -0.796528 0.170989 -4.658
## block2:BBBB 0.314092 0.117970 2.662
## block4:BBBB -0.246985 0.117668 -2.099
## block2:AFMN 0.623770 0.122010 5.112
## block4:AFMN -0.232500 0.121342 -1.916
## BBBB:AFMN -0.002289 0.271234 -0.008
## block2:LabelE -0.043551 0.122072 -0.357
## block4:LabelE -0.101170 0.121639 -0.832
## BBBB:LabelE -0.189121 0.118295 -1.599
## AFMN:LabelE 0.271842 0.122037 2.228
## block2:BBBB:AFMN -0.381098 0.170163 -2.240
## block4:BBBB:AFMN 0.348422 0.169413 2.057
## block2:BBBB:LabelE -0.093649 0.167005 -0.561
## block4:BBBB:LabelE 0.328207 0.166852 1.967
## block2:AFMN:LabelE 0.130504 0.172402 0.757
## block4:AFMN:LabelE 0.329845 0.172193 1.916
## BBBB:AFMN:LabelE -0.168344 0.169557 -0.993
## block2:BBBB:AFMN:LabelE -0.032519 0.240300 -0.135
## block4:BBBB:AFMN:LabelE -0.489136 0.239892 -2.039##
## Correlation matrix not shown by default, as p = 25 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need itlibrary(lme4)
f2barkallbl <- lmer(F2Bark ~ block * BB * AF* Label + gender + (1|subj) + (1|word), data = df)
summary(f2barkallbl)## Linear mixed model fit by REML ['lmerMod']
## Formula: F2Bark ~ block * BB * AF * Label + gender + (1 | subj) + (1 |
## word)
## Data: df
##
## REML criterion at convergence: 5195
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.4683 -0.5559 0.0047 0.6209 3.7439
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 0.13125 0.3623
## word (Intercept) 0.09871 0.3142
## Residual 0.33280 0.5769
## Number of obs: 2847, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 13.53089 0.21657 62.477
## block2 -0.03860 0.07609 -0.507
## block4 -0.16534 0.07559 -2.187
## BBBB -0.23626 0.15582 -1.516
## AFMN 0.02184 0.16261 0.134
## LabelE -0.75914 0.26768 -2.836
## genderM -0.88032 0.13936 -6.317
## block2:BBBB -2.56574 0.10427 -24.606
## block4:BBBB 0.06241 0.10401 0.600
## block2:AFMN -0.20012 0.10784 -1.856
## block4:AFMN 0.09089 0.10725 0.847
## BBBB:AFMN 0.31109 0.22419 1.388
## block2:LabelE -0.09628 0.10790 -0.892
## block4:LabelE 0.04710 0.10751 0.438
## BBBB:LabelE 0.08442 0.10456 0.807
## AFMN:LabelE -0.07222 0.10787 -0.670
## block2:BBBB:AFMN -0.21673 0.15040 -1.441
## block4:BBBB:AFMN -0.16465 0.14974 -1.100
## block2:BBBB:LabelE 0.73355 0.14761 4.969
## block4:BBBB:LabelE 0.03067 0.14748 0.208
## block2:AFMN:LabelE 0.17939 0.15238 1.177
## block4:AFMN:LabelE 0.03849 0.15220 0.253
## BBBB:AFMN:LabelE -0.10261 0.14987 -0.685
## block2:BBBB:AFMN:LabelE -0.12610 0.21240 -0.594
## block4:BBBB:AFMN:LabelE -0.06195 0.21204 -0.292##
## Correlation matrix not shown by default, as p = 25 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need itRidge plots
ridge plots for baseline description
#in hertz
library(ggplot2)
library(ggridges)
library(tidyr)##
## Attaching package: 'tidyr'## The following objects are masked from 'package:Matrix':
##
## expand, pack, unpack#define color-blindness-friendly colors
vowelcolor <- c("#0072B2", "#CC79A7")
ridgeBaselinepivot <- df |> filter(block == "1") |>
pivot_longer(
cols = c("F1Hz","F2Hz"),
names_to = "formant",
values_to = "Hertz",
)
#head(ridgeBaselinepivot)
ggplot(ridgeBaselinepivot, aes(x = Hertz, y = formant, fill = Label)) +
geom_density_ridges(alpha = 0.6, scale = 1) +
theme_ridges() +
scale_fill_discrete(breaks = c ("F1Hertz","F2Hertz")) +
scale_fill_manual(values = vowelcolor)+
ggtitle("formant values at pretest, pooled data")## Scale for fill is already present.
## Adding another scale for fill, which will replace the existing scale.## Picking joint bandwidth of 37.6
#ggsave("ridgeBlock1pooled.png", plot = last_plot(), device = NULL, scale = 1,dpi = 600)
#alternatively
ridgebaseline <- df |> filter(block == "1")
colors <- c("F1Hz" = "#0072B2", "F2Hz" = "#CC79A7")
ggplot(ridgebaseline, aes(y = Label)) +
geom_density_ridges(aes(x = F1Hz, fill = "F1Hz"), alpha = 0.6, scale = 1) +
geom_density_ridges(aes(x = F2Hz, fill = "F2Hz"), alpha = 0.6, scale = 1)+
#scale_fill_manual(values = "#0072B2")+
theme_ridges() +labs(x = "Hz", y = "phoneme", fill = "Legend")#+## Picking joint bandwidth of 21.2## Picking joint bandwidth of 54.1
#ggtitle("formant values at pretest, pooled data")
#ggsave("ridgeBlock1pooled.png", plot = last_plot(), device = NULL, scale = 1,dpi = 600)ridge plots for RQ1
#in hertz
#AF, F1
afdf <- df |> filter((AF == "AF" & block == "1") | (AF == "AF" & block == "2"))
ggplot(afdf, aes(x = F1Hz, y = block, fill = BB, linetype = Label)) +
geom_density_ridges(alpha = 0.6, scale = 1.2) +
theme_ridges() +
#scale_fill_discrete(breaks = c ("F1Hertz","F2Hertz")) +
scale_fill_manual(values = vowelcolor)## Picking joint bandwidth of 29.2
#+ggtitle("F1 in two vowels between groups in each test block, with regular auditory feedback")
#ggsave("ridgeF1BlockXBB*vowel_AF.png", plot = last_plot(), device = NULL, scale = 1,dpi = 600)
#AF, F2
ggplot(afdf, aes(x = F2Hz, y = block, fill = BB, linetype = Label)) +
geom_density_ridges(alpha = 0.6, scale = 1.2) +
theme_ridges() +
#scale_fill_discrete(breaks = c ("F1Hertz","F2Hertz")) +
scale_fill_manual(values = vowelcolor)## Picking joint bandwidth of 71.2
#+ggtitle("formant values in two vowels between groups in each test block, with regular auditory feedback")
#ggsave("ridgeF2BlockXBB*vowel_AF.png", plot = last_plot(), device = NULL, scale = 1,dpi = 600)
# MN F1
mndf <- df |> filter(AF == "MN")
ggplot(mndf, aes(x = F1Hz, y = block, fill = BB, linetype = Label)) +
geom_density_ridges(alpha = 0.6, scale = 1.2) +
theme_ridges() +
#scale_fill_discrete(breaks = c ("F1Hertz","F2Hertz")) +
scale_fill_manual(values = vowelcolor)## Picking joint bandwidth of 27.3
#+ggtitle("formant values in two vowels between groups in each test block, with masking noise")
#ggsave("ridgeF1BlockXBB*vowel_MN.png", plot = last_plot(), device = NULL, scale = 1,dpi = 600)
#MN F2
ggplot(mndf, aes(x = F2Hz, y = block, fill = BB, linetype = Label)) +
geom_density_ridges(alpha = 0.6, scale = 1.2) +
theme_ridges() +
#scale_fill_discrete(breaks = c ("F1Hertz","F2Hertz")) +
scale_fill_manual(values = vowelcolor)## Picking joint bandwidth of 65.2
#+ggtitle("formant values in two vowels between groups in each test block, with masking noise")
#ggsave("ridgeF2BlockXBB*vowel_MN.png", plot = last_plot(), device = NULL, scale = 1,dpi = 600)after the meeting on Mar 18, WE ARE SEPARATING THE VOWELS and removing block 4
#AF, F1 vowel E
rq1e <- df |> filter((AF == "AF" & Label == "E" & block == "1") | (AF == "AF" & Label == "E" & block == "2"))
p1 <- ggplot(rq1e, aes(x = F1Hz, y = block, fill = BB)) +
geom_density_ridges(alpha = 0.6, scale = 1.0) +
theme_ridges() +
#scale_fill_discrete(breaks = c ("F1Hertz","F2Hertz")) +
scale_fill_manual(values = vowelcolor)+
ggtitle("vowel E")
#AF, F1 vowel I
rq1e <- df |> filter((AF == "AF" & Label == "I" & block == "1") | (AF == "AF" & Label == "I" & block == "2"))
p2 <- ggplot(rq1e, aes(x = F1Hz, y = block, fill = BB)) +
geom_density_ridges(alpha = 0.6, scale = 1.0) +
theme_ridges() +
#scale_fill_discrete(breaks = c ("F1Hertz","F2Hertz")) +
scale_fill_manual(values = vowelcolor)+
ggtitle("vowel I")
p1 + p2## Picking joint bandwidth of 35.5## Picking joint bandwidth of 22.9
ridge plots for RQ2
#BB*AF at block 2 because that is the only significant block
#F1
ggplot(df, aes(x = F1Hz, y = AF, fill = BB, linetype = Label)) +
geom_density_ridges(alpha = 0.6, scale = 1.2) +
theme_ridges() +
scale_fill_manual(values = vowelcolor)## Picking joint bandwidth of 23.6
#+ggtitle("formant values in two vowels between groups in each test block, with masking noise")
#ggsave("ridgeF1Block*BB_block2.png", plot = last_plot(), device = NULL, scale = 1,dpi = 600)
#hyper articulation in the MN condition because Lombard?
#F2
ggplot(df, aes(x = F2Hz, y = AF, fill = BB, linetype = Label)) +
geom_density_ridges(alpha = 0.6, scale = 1.2) +
theme_ridges() +
#scale_fill_discrete(breaks = c ("F1Hertz","F2Hertz")) +
scale_fill_manual(values = vowelcolor)## Picking joint bandwidth of 77.1
#+ggtitle("formant values in two vowels between groups in each test block, with masking noise")
#ggsave("ridgeF2Block*BB_block2.png", plot = last_plot(), device = NULL, scale = 1,dpi = 600)ridge plots for RQ3
#AF
#reshaping data
library(tidyr)
ridgedfaf <- df |> filter(block == "4", AF == "AF") |>
pivot_longer(
cols = c("F1Hz","F2Hz"),
names_to = "formant",
values_to = "Hertz",
)
head(ridgedfaf)## # A tibble: 6 × 14
## fileName subj word block seq Label F0Hz F1Bark F2Bark gender BB AF
## <chr> <chr> <chr> <fct> <dbl> <fct> <dbl> <dbl> <dbl> <fct> <fct> <fct>
## 1 subj_10_… 10 kigg… 4 13 I 216 4 14 F BB AF
## 2 subj_10_… 10 kigg… 4 13 I 216 4 14 F BB AF
## 3 subj_10_… 10 kigg… 4 17 I 212 4 14 F BB AF
## 4 subj_10_… 10 kigg… 4 17 I 212 4 14 F BB AF
## 5 subj_10_… 10 kigg… 4 26 I 223 4 14 F BB AF
## 6 subj_10_… 10 kigg… 4 26 I 223 4 14 F BB AF
## # ℹ 2 more variables: formant <chr>, Hertz <dbl>ggplot(ridgedfaf, aes(x = Hertz, y = formant, fill = BB, linetype = Label)) +
geom_density_ridges(alpha = 0.6) +
theme_ridges() +
scale_fill_discrete(breaks = c ("F1Hertz","F2Hertz")) +
scale_fill_manual(values = vowelcolor)+
ggtitle("formant values at block 4, for AF group")## Scale for fill is already present.
## Adding another scale for fill, which will replace the existing scale.
## Picking joint bandwidth of 47.3
ggsave("ridgeBlock4AF.png", plot = last_plot(), device = NULL, scale = 1,dpi = 600)## Saving 7 x 5 in image
## Picking joint bandwidth of 47.3#MN
#reshaping data
library(tidyr)
ridgedfaf <- df |> filter(block == "4", AF == "MN") |>
pivot_longer(
cols = c("F1Hz","F2Hz"),
names_to = "formant",
values_to = "Hertz",
)
head(ridgedfaf)## # A tibble: 6 × 14
## fileName subj word block seq Label F0Hz F1Bark F2Bark gender BB AF
## <chr> <chr> <chr> <fct> <dbl> <fct> <dbl> <dbl> <dbl> <fct> <fct> <fct>
## 1 subj_11_… 11 kigg… 4 12 I 230 4 14 F NB MN
## 2 subj_11_… 11 kigg… 4 12 I 230 4 14 F NB MN
## 3 subj_11_… 11 kigg… 4 41 I 227 4 14 F NB MN
## 4 subj_11_… 11 kigg… 4 41 I 227 4 14 F NB MN
## 5 subj_11_… 11 kigg… 4 6 I 224 4 14 F NB MN
## 6 subj_11_… 11 kigg… 4 6 I 224 4 14 F NB MN
## # ℹ 2 more variables: formant <chr>, Hertz <dbl>ggplot(ridgedfaf, aes(x = Hertz, y = formant, fill = BB)) +
geom_density_ridges(alpha = 0.6) +
theme_ridges() +
scale_fill_discrete(breaks = c ("F1Hertz","F2Hertz")) +
scale_fill_manual(values = vowelcolor)+
ggtitle("formant values at block 4, for MN group")## Scale for fill is already present.
## Adding another scale for fill, which will replace the existing scale.
## Picking joint bandwidth of 49.3
ggsave("ridgeBlock4MN.png", plot = last_plot(), device = NULL, scale = 1,dpi = 600)## Saving 7 x 5 in image
## Picking joint bandwidth of 49.3#reshaping data
library(tidyr)
ridgedf <- df |> filter(block == "2", AF == "AF") |>
pivot_longer(
cols = c("F1Bark","F2Bark"),
names_to = "formant",
values_to = "Bark",
)
head(ridgedf)## # A tibble: 6 × 14
## fileName subj word block seq Label F0Hz F1Hz F2Hz gender BB AF
## <chr> <chr> <chr> <fct> <dbl> <fct> <dbl> <dbl> <dbl> <fct> <fct> <fct>
## 1 subj_10_ki… 10 kigg… 2 28 I 245 504 1326 F BB AF
## 2 subj_10_ki… 10 kigg… 2 28 I 245 504 1326 F BB AF
## 3 subj_10_ki… 10 kigg… 2 52 I 236 460 953 F BB AF
## 4 subj_10_ki… 10 kigg… 2 52 I 236 460 953 F BB AF
## 5 subj_10_pi… 10 piff… 2 10 I 234 379 1099 F BB AF
## 6 subj_10_pi… 10 piff… 2 10 I 234 379 1099 F BB AF
## # ℹ 2 more variables: formant <chr>, Bark <dbl>library(ggplot2)
library(ggridges)
#define color-blindness-friendly colors
color <- c("#0072B2", "#CC79A7")
ggplot(ridgedf, aes(x = Bark, y = formant, fill = BB)) +
geom_density_ridges(alpha = 0.6) +
theme_ridges() +
scale_fill_discrete(breaks = c ("F1Bark","F2Bark")) +
scale_fill_manual(values = color)+
ggtitle("formant values at block 2, for AF condition")## Scale for fill is already present.
## Adding another scale for fill, which will replace the existing scale.
## Picking joint bandwidth of 0.251
#ggsave("ridgeBlock2AFBark.png", plot = last_plot(), device = NULL, scale = 1,dpi = 600)#reshaping data
library(tidyr)
ridgedf <- df |> filter(block == "2", AF == "MN") |>
pivot_longer(
cols = c("F1Hz","F2Hz"),
names_to = "formant",
values_to = "Hertz",
)
head(ridgedf)## # A tibble: 6 × 14
## fileName subj word block seq Label F0Hz F1Bark F2Bark gender BB AF
## <chr> <chr> <chr> <fct> <dbl> <fct> <dbl> <dbl> <dbl> <fct> <fct> <fct>
## 1 subj_11_… 11 kigg… 2 13 I 251 5 14 F NB MN
## 2 subj_11_… 11 kigg… 2 13 I 251 5 14 F NB MN
## 3 subj_11_… 11 kigg… 2 48 I 251 5 14 F NB MN
## 4 subj_11_… 11 kigg… 2 48 I 251 5 14 F NB MN
## 5 subj_11_… 11 kigg… 2 51 I 247 5 14 F NB MN
## 6 subj_11_… 11 kigg… 2 51 I 247 5 14 F NB MN
## # ℹ 2 more variables: formant <chr>, Hertz <dbl>library(ggplot2)
library(ggridges)
#define color-blindness-friendly colors
color <- c("#0072B2", "#CC79A7")
ggplot(ridgedf, aes(x = Hertz, y = formant, fill = BB)) +
geom_density_ridges(alpha = 0.6) +
theme_ridges() +
scale_fill_discrete(breaks = c ("F1Hz","F2Hz")) +
scale_fill_manual(values = color)+
ggtitle("formant values at block 2, for MN condition")## Scale for fill is already present.
## Adding another scale for fill, which will replace the existing scale.
## Picking joint bandwidth of 53.6
#ggsave("ridgeBlock2MN.png", plot = last_plot(), device = NULL, scale = 1,dpi = 600)#reshaping data
library(tidyr)
ridge4af <- df |> filter(block == "4", AF == "AF") |>
pivot_longer(
cols = c("F1Hz","F2Hz"),
names_to = "formant",
values_to = "Hertz",
)
head(ridgedf)## # A tibble: 6 × 14
## fileName subj word block seq Label F0Hz F1Bark F2Bark gender BB AF
## <chr> <chr> <chr> <fct> <dbl> <fct> <dbl> <dbl> <dbl> <fct> <fct> <fct>
## 1 subj_11_… 11 kigg… 2 13 I 251 5 14 F NB MN
## 2 subj_11_… 11 kigg… 2 13 I 251 5 14 F NB MN
## 3 subj_11_… 11 kigg… 2 48 I 251 5 14 F NB MN
## 4 subj_11_… 11 kigg… 2 48 I 251 5 14 F NB MN
## 5 subj_11_… 11 kigg… 2 51 I 247 5 14 F NB MN
## 6 subj_11_… 11 kigg… 2 51 I 247 5 14 F NB MN
## # ℹ 2 more variables: formant <chr>, Hertz <dbl>library(ggplot2)
library(ggridges)
#define color-blindness-friendly colors
color <- c("#0072B2", "#CC79A7")
ggplot(ridge4af, aes(x = Hertz, y = formant, fill = BB)) +
geom_density_ridges(alpha = 0.6) +
theme_ridges() +
scale_fill_discrete(breaks = c ("F1Hz","F2Hz")) +
ggtitle("formant values at block 4, for AF condition")## Picking joint bandwidth of 51
scale_fill_manual(values = color)## <ggproto object: Class ScaleDiscrete, Scale, gg>
## aesthetics: fill
## axis_order: function
## break_info: function
## break_positions: function
## breaks: waiver
## call: call
## clone: function
## dimension: function
## drop: TRUE
## expand: waiver
## get_breaks: function
## get_breaks_minor: function
## get_labels: function
## get_limits: function
## get_transformation: function
## guide: legend
## is_discrete: function
## is_empty: function
## labels: waiver
## limits: NULL
## make_sec_title: function
## make_title: function
## map: function
## map_df: function
## n.breaks.cache: NULL
## na.translate: TRUE
## na.value: grey50
## name: waiver
## palette: function
## palette.cache: NULL
## position: left
## range: environment
## rescale: function
## reset: function
## train: function
## train_df: function
## transform: function
## transform_df: function
## super: <ggproto object: Class ScaleDiscrete, Scale, gg>#ggsave("ridgeBlock4AF.png", plot = last_plot(), device = NULL, scale = 1,dpi = 600)#in bark
library(tidyr)
ridge4af <- df |> filter(block == "4", AF == "AF") |>
pivot_longer(
cols = c("F1Bark","F2Bark"),
names_to = "formant",
values_to = "Bark",
)
head(ridgedf)## # A tibble: 6 × 14
## fileName subj word block seq Label F0Hz F1Bark F2Bark gender BB AF
## <chr> <chr> <chr> <fct> <dbl> <fct> <dbl> <dbl> <dbl> <fct> <fct> <fct>
## 1 subj_11_… 11 kigg… 2 13 I 251 5 14 F NB MN
## 2 subj_11_… 11 kigg… 2 13 I 251 5 14 F NB MN
## 3 subj_11_… 11 kigg… 2 48 I 251 5 14 F NB MN
## 4 subj_11_… 11 kigg… 2 48 I 251 5 14 F NB MN
## 5 subj_11_… 11 kigg… 2 51 I 247 5 14 F NB MN
## 6 subj_11_… 11 kigg… 2 51 I 247 5 14 F NB MN
## # ℹ 2 more variables: formant <chr>, Hertz <dbl>library(ggplot2)
library(ggridges)
#define color-blindness-friendly colors
color <- c("#0072B2", "#CC79A7")
ggplot(ridge4af, aes(x = Bark, y = formant, fill = BB)) +
geom_density_ridges(alpha = 0.6) +
theme_ridges() +
scale_fill_discrete(breaks = c ("F1Bark","F2Bark")) +
ggtitle("formant values at block 4, for AF condition")## Picking joint bandwidth of 0.247
scale_fill_manual(values = color)## <ggproto object: Class ScaleDiscrete, Scale, gg>
## aesthetics: fill
## axis_order: function
## break_info: function
## break_positions: function
## breaks: waiver
## call: call
## clone: function
## dimension: function
## drop: TRUE
## expand: waiver
## get_breaks: function
## get_breaks_minor: function
## get_labels: function
## get_limits: function
## get_transformation: function
## guide: legend
## is_discrete: function
## is_empty: function
## labels: waiver
## limits: NULL
## make_sec_title: function
## make_title: function
## map: function
## map_df: function
## n.breaks.cache: NULL
## na.translate: TRUE
## na.value: grey50
## name: waiver
## palette: function
## palette.cache: NULL
## position: left
## range: environment
## rescale: function
## reset: function
## train: function
## train_df: function
## transform: function
## transform_df: function
## super: <ggproto object: Class ScaleDiscrete, Scale, gg>#reshaping data
library(tidyr)
ridge4mn <- df |> filter(block == "4", AF == "MN") |>
pivot_longer(
cols = c("F1Hz","F2Hz"),
names_to = "formant",
values_to = "Hertz",
)
head(ridgedf)## # A tibble: 6 × 14
## fileName subj word block seq Label F0Hz F1Bark F2Bark gender BB AF
## <chr> <chr> <chr> <fct> <dbl> <fct> <dbl> <dbl> <dbl> <fct> <fct> <fct>
## 1 subj_11_… 11 kigg… 2 13 I 251 5 14 F NB MN
## 2 subj_11_… 11 kigg… 2 13 I 251 5 14 F NB MN
## 3 subj_11_… 11 kigg… 2 48 I 251 5 14 F NB MN
## 4 subj_11_… 11 kigg… 2 48 I 251 5 14 F NB MN
## 5 subj_11_… 11 kigg… 2 51 I 247 5 14 F NB MN
## 6 subj_11_… 11 kigg… 2 51 I 247 5 14 F NB MN
## # ℹ 2 more variables: formant <chr>, Hertz <dbl>library(ggplot2)
library(ggridges)
#define color-blindness-friendly colors
color <- c("#0072B2", "#CC79A7")
ggplot(ridge4mn, aes(x = Hertz, y = formant, fill = BB)) +
geom_density_ridges(alpha = 0.6) +
theme_ridges() +
scale_fill_discrete(breaks = c ("F1Hz","F2Hz")) +
ggtitle("formant values at block 4, for MN condition")## Picking joint bandwidth of 49.3
scale_fill_manual(values = color)## <ggproto object: Class ScaleDiscrete, Scale, gg>
## aesthetics: fill
## axis_order: function
## break_info: function
## break_positions: function
## breaks: waiver
## call: call
## clone: function
## dimension: function
## drop: TRUE
## expand: waiver
## get_breaks: function
## get_breaks_minor: function
## get_labels: function
## get_limits: function
## get_transformation: function
## guide: legend
## is_discrete: function
## is_empty: function
## labels: waiver
## limits: NULL
## make_sec_title: function
## make_title: function
## map: function
## map_df: function
## n.breaks.cache: NULL
## na.translate: TRUE
## na.value: grey50
## name: waiver
## palette: function
## palette.cache: NULL
## position: left
## range: environment
## rescale: function
## reset: function
## train: function
## train_df: function
## transform: function
## transform_df: function
## super: <ggproto object: Class ScaleDiscrete, Scale, gg>#ggsave("ridgeBlock4MN.png", plot = last_plot(), device = NULL, scale = 1,dpi = 600)#reshaping data
library(tidyr)
ridge4mn <- df |> filter(block == "4", AF == "MN") |>
pivot_longer(
cols = c("F1Bark","F2Bark"),
names_to = "formant",
values_to = "Bark",
)
head(ridgedf)## # A tibble: 6 × 14
## fileName subj word block seq Label F0Hz F1Bark F2Bark gender BB AF
## <chr> <chr> <chr> <fct> <dbl> <fct> <dbl> <dbl> <dbl> <fct> <fct> <fct>
## 1 subj_11_… 11 kigg… 2 13 I 251 5 14 F NB MN
## 2 subj_11_… 11 kigg… 2 13 I 251 5 14 F NB MN
## 3 subj_11_… 11 kigg… 2 48 I 251 5 14 F NB MN
## 4 subj_11_… 11 kigg… 2 48 I 251 5 14 F NB MN
## 5 subj_11_… 11 kigg… 2 51 I 247 5 14 F NB MN
## 6 subj_11_… 11 kigg… 2 51 I 247 5 14 F NB MN
## # ℹ 2 more variables: formant <chr>, Hertz <dbl>library(ggplot2)
library(ggridges)
#define color-blindness-friendly colors
color <- c("#0072B2", "#CC79A7")
ggplot(ridge4mn, aes(x = Bark, y = formant, fill = BB)) +
geom_density_ridges(alpha = 0.6) +
theme_ridges() +
scale_fill_discrete(breaks = c ("F1Bark","F2Bark")) +
ggtitle("formant values at block 4, for MN condition")## Picking joint bandwidth of 0.242
scale_fill_manual(values = color)## <ggproto object: Class ScaleDiscrete, Scale, gg>
## aesthetics: fill
## axis_order: function
## break_info: function
## break_positions: function
## breaks: waiver
## call: call
## clone: function
## dimension: function
## drop: TRUE
## expand: waiver
## get_breaks: function
## get_breaks_minor: function
## get_labels: function
## get_limits: function
## get_transformation: function
## guide: legend
## is_discrete: function
## is_empty: function
## labels: waiver
## limits: NULL
## make_sec_title: function
## make_title: function
## map: function
## map_df: function
## n.breaks.cache: NULL
## na.translate: TRUE
## na.value: grey50
## name: waiver
## palette: function
## palette.cache: NULL
## position: left
## range: environment
## rescale: function
## reset: function
## train: function
## train_df: function
## transform: function
## transform_df: function
## super: <ggproto object: Class ScaleDiscrete, Scale, gg>====== old version below =====
#filter for the pre and post conditions
prepost <- db |> filter(block == "Pre" | block == "Post")
f1hzprepost <- lmer(F1Hz ~ block * BB * block* Label + gender + (1|subj) + (1|word), data = prepost)
summary(f1hzprepost)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Hz ~ block * BB * block * Label + gender + (1 | subj) + (1 |
## word)
## Data: prepost
##
## REML criterion at convergence: 21501.3
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.6421 -0.5144 0.0187 0.5142 18.3661
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 2348 48.46
## word (Intercept) 284 16.85
## Residual 4483 66.95
## Number of obs: 1902, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 580.827 14.400 40.336
## blockPre -2.427 6.040 -0.402
## BBNB 21.639 14.570 1.485
## LabelI -142.152 15.037 -9.454
## genderM -84.508 18.188 -4.646
## blockPre:BBNB 7.080 8.703 0.813
## blockPre:LabelI 8.183 8.550 0.957
## BBNB:LabelI -19.923 8.695 -2.291
## blockPre:BBNB:LabelI -13.845 12.293 -1.126
##
## Correlation of Fixed Effects:
## (Intr) blckPr BBNB LabelI gendrM blP:BBNB blP:LI BBNB:L
## blockPre -0.211
## BBNB -0.494 0.208
## LabelI -0.522 0.202 0.084
## genderM -0.225 0.000 0.030 0.000
## blckPr:BBNB 0.146 -0.694 -0.299 -0.140 0.001
## blckPr:LblI 0.149 -0.706 -0.147 -0.286 0.000 0.490
## BBNB:LabelI 0.146 -0.349 -0.299 -0.281 0.000 0.501 0.495
## blP:BBNB:LI -0.103 0.491 0.212 0.199 -0.001 -0.708 -0.696 -0.707f2hzprepost <- lmer(F2Hz ~ block * BB * block* Label + gender + (1|subj) + (1|word), data = prepost)
summary(f2hzprepost)## Linear mixed model fit by REML ['lmerMod']
## Formula: F2Hz ~ block * BB * block * Label + gender + (1 | subj) + (1 |
## word)
## Data: prepost
##
## REML criterion at convergence: 23423.6
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -6.2328 -0.5602 -0.0181 0.5161 6.4825
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 13382 115.7
## word (Intercept) 14516 120.5
## Residual 12062 109.8
## Number of obs: 1902, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 1920.967 73.649 26.083
## blockPre 34.290 9.907 3.461
## BBNB 24.180 33.102 0.730
## LabelI 214.029 98.875 2.165
## genderM -254.596 42.814 -5.947
## blockPre:BBNB -27.480 14.277 -1.925
## blockPre:LabelI 1.544 14.025 0.110
## BBNB:LabelI 12.458 14.263 0.873
## blockPre:BBNB:LabelI 24.773 20.165 1.229
##
## Correlation of Fixed Effects:
## (Intr) blckPr BBNB LabelI gendrM blP:BBNB blP:LI BBNB:L
## blockPre -0.068
## BBNB -0.220 0.150
## LabelI -0.671 0.050 0.015
## genderM -0.104 0.000 0.032 0.000
## blckPr:BBNB 0.047 -0.694 -0.216 -0.035 0.001
## blckPr:LblI 0.048 -0.706 -0.106 -0.071 0.000 0.490
## BBNB:LabelI 0.047 -0.349 -0.216 -0.070 0.000 0.501 0.495
## blP:BBNB:LI -0.033 0.491 0.153 0.050 -0.001 -0.708 -0.696 -0.707#filter for MN1 and AF1, comprare with pretest
bl1 <- db |> filter(block == "Pre" | block == "MN1" | block == "AF1")
af1subj <- c("2", "3", "5", "6", "8", "9", "10", "12", "15", "17", "19", "26", "30", "31", "33", "35", "38", "40", "42", "43", "46", "48", "50", "51", "52", "53", "54", "56", "58", "60")
bl1af <- bl1 |> mutate(
AF = case_when(
subj %in% af1subj ~ "AF",
TRUE ~ "MN")) |> mutate(
block = case_when(
block == "AF1" ~ "2",
block == "MN1" ~ "2",
block == "Pre" ~ "1")
)
#repeating contrast coding so as to not change results when rerunning this block after running the next block
#contrast coding
bl1af$BB <- as.factor(bl1af$BB)
bl1af$BB <- relevel(bl1af$BB, ref = "NB")
bl1af$Label <- as.factor(bl1af$Label)
bl1af$Label <- relevel(bl1af$Label, ref = "I")
bl1af$block <- as.factor(bl1af$block)
bl1af$block <- relevel(bl1af$block, ref = "1")
bl1af$gender <- as.factor(bl1af$gender)
bl1af$gender <- relevel(bl1af$gender, ref = "F")
f1hzbl1af <- lmer(F1Hz ~ block * BB * AF* Label + gender + (1|subj) + (1|word), data = bl1af)
summary(f1hzbl1af)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Hz ~ block * BB * AF * Label + gender + (1 | subj) + (1 | word)
## Data: bl1af
##
## REML criterion at convergence: 21727.1
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.8859 -0.5101 0.0280 0.4989 9.6281
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 3999.45 63.241
## word (Intercept) 93.05 9.646
## Residual 5353.21 73.166
## Number of obs: 1897, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 452.0889 20.3977 22.164
## block2 0.5538 9.6518 0.057
## BBBB 9.3619 25.7308 0.364
## AFMN -19.4367 26.8760 -0.723
## LabelE 147.1272 12.4950 11.775
## genderM -102.5361 24.2120 -4.235
## block2:BBBB 36.7324 13.2256 2.777
## block2:AFMN 63.0005 13.6783 4.606
## BBBB:AFMN -10.0344 37.0205 -0.271
## block2:LabelE -7.7660 13.6887 -0.567
## BBBB:LabelE -20.4306 13.2645 -1.540
## AFMN:LabelE 39.5689 13.6825 2.892
## block2:BBBB:AFMN -24.7403 19.0799 -1.297
## block2:BBBB:LabelE -9.9069 18.7256 -0.529
## block2:AFMN:LabelE 35.3210 19.3298 1.827
## BBBB:AFMN:LabelE -24.0924 19.0102 -1.267
## block2:BBBB:AFMN:LabelE -27.6713 26.9425 -1.027##
## Correlation matrix not shown by default, as p = 17 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need itf2hzbl1af <- lmer(F2Hz ~ block * BB * AF* Label + gender + (1|subj) + (1|word), data = bl1af)
summary(f2hzbl1af)## Linear mixed model fit by REML ['lmerMod']
## Formula: F2Hz ~ block * BB * AF * Label + gender + (1 | subj) + (1 | word)
## Data: bl1af
##
## REML criterion at convergence: 24145.7
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.1706 -0.5581 -0.0062 0.5393 5.3927
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 11225 105.95
## word (Intercept) 7947 89.14
## Residual 19378 139.20
## Number of obs: 1897, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 2198.521 61.385 35.816
## block2 -9.033 18.363 -0.492
## BBBB -79.790 43.920 -1.817
## AFMN 6.701 45.863 0.146
## LabelE -255.274 75.089 -3.400
## genderM -232.921 40.781 -5.712
## block2:BBBB -652.633 25.163 -25.936
## block2:AFMN -55.221 26.024 -2.122
## BBBB:AFMN 97.900 63.190 1.549
## block2:LabelE -16.157 26.044 -0.620
## BBBB:LabelE 64.472 25.237 2.555
## AFMN:LabelE 5.340 26.032 0.205
## block2:BBBB:AFMN -85.842 36.301 -2.365
## block2:BBBB:LabelE 192.300 35.627 5.398
## block2:AFMN:LabelE 37.361 36.776 1.016
## BBBB:AFMN:LabelE -61.094 36.168 -1.689
## block2:BBBB:AFMN:LabelE 8.542 51.260 0.167##
## Correlation matrix not shown by default, as p = 17 > 12.
## Use print(x, correlation=TRUE) or
## vcov(x) if you need itdbaf<- db|> filter(block == "AF1" | block == "Pre") |> filter(Label == "E")
ggplot(data = dbaf, aes(x = F2Hz, y = F1Hz, color = block))+
geom_point(alpha = 0.5)
Interpretation: F1 model:
blockMN1 53.658 9.036 5.938
MN group has 53.758Hz higher F1 at test compared to pretest (whereas AF group did not show difference) -> Lombard?
LabelE 167.017 10.450 15.983
/E/ is 167.017Hz higher in F1 than /I/(at pretest).
genderM -98.731 23.301 -4.237
Male speakers’ F1 is 98.731Hz lower than female speakers (at pretest)
blockAF1:BBBB 31.336 12.370 2.533
The F1 difference between NB and BB group is 31Hz bigger at AF1 than at pretest (note that at pretest the BB group did not have the bite block in)
blockMN1:LabelE 47.236 11.855 3.985
The F1 difference between /E/ and /I/ is bigger at MN1 than at pretest (auditory feedback at pretest is AF)
BBBB:LabelE -33.046 9.517 -3.472
The F1 difference between /E/ and /I/ is smaller in the BB group than in the NB group at pretest
blockMN1:BBBB:LabelE -49.059 16.768 -2.926
?
#splitting up MN1 and AF1, compare with pretest respectively
af <- db |> filter(block == "Pre" | block == "AF1")
mn <- db |> filter(block == "Pre" | block == "MN1")
#contrast coding
af$BB <- as.factor(af$BB)
af$BB <- relevel(af$BB, ref = "NB")
af$Label <- as.factor(af$Label)
af$Label <- relevel(af$Label, ref = "I")
af$block <- as.factor(af$block)
af$block <- relevel(af$block, ref = "Pre")
af$gender <- as.factor(af$gender)
af$gender <- relevel(af$gender, ref = "F")
f1hzaf <- lmer(F1Hz ~ block * BB * Label + gender + (1|subj) + (1|word), data = af)
summary(f1hzaf)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Hz ~ block * BB * Label + gender + (1 | subj) + (1 | word)
## Data: af
##
## REML criterion at convergence: 16221.6
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.2819 -0.5589 0.0282 0.5470 11.0900
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 3238 56.90
## word (Intercept) 204 14.28
## Residual 4069 63.79
## Number of obs: 1445, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 440.852 14.865 29.657
## blockAF1 10.329 7.846 1.316
## BBBB 5.255 16.573 0.317
## LabelE 167.312 13.098 12.774
## genderM -93.417 21.272 -4.392
## blockAF1:BBBB 31.225 10.767 2.900
## blockAF1:LabelE -27.850 10.317 -2.699
## BBBB:LabelE -33.366 8.280 -4.030
## blockAF1:BBBB:LabelE 2.991 14.189 0.211
##
## Correlation of Fixed Effects:
## (Intr) blcAF1 BBBB LabelE gendrM blAF1:BBBB bAF1:L BBBB:L
## blockAF1 -0.148
## BBBB -0.571 0.136
## LabelE -0.440 0.170 0.082
## genderM -0.220 -0.013 -0.031 0.001
## blcAF1:BBBB 0.111 -0.729 -0.191 -0.124 -0.003
## blckAF1:LbE 0.116 -0.657 -0.103 -0.263 -0.003 0.479
## BBBB:LabelE 0.144 -0.269 -0.249 -0.328 -0.001 0.381 0.417
## bAF1:BBBB:L -0.084 0.478 0.145 0.192 0.003 -0.660 -0.727 -0.584mn$BB <- as.factor(mn$BB)
mn$BB <- relevel(mn$BB, ref = "NB")
mn$Label <- as.factor(mn$Label)
mn$Label <- relevel(mn$Label, ref = "I")
mn$block <- as.factor(mn$block)
mn$block <- relevel(mn$block, ref = "Pre")
mn$gender <- as.factor(mn$gender)
mn$gender <- relevel(mn$gender, ref = "F")
f1hzmn <- lmer(F1Hz ~ block * BB * block* Label + gender + (1|subj) + (1|word), data = mn)
summary(f1hzmn)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Hz ~ block * BB * block * Label + gender + (1 | subj) + (1 |
## word)
## Data: mn
##
## REML criterion at convergence: 16195.6
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.7920 -0.4615 0.0561 0.4755 9.1062
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 3457.4 58.80
## word (Intercept) 117.1 10.82
## Residual 5611.9 74.91
## Number of obs: 1404, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 440.964 14.432 30.554
## blockMN1 53.786 9.226 5.830
## BBBB 4.803 17.433 0.276
## LabelE 167.281 11.276 14.835
## genderM -93.159 22.364 -4.166
## blockMN1:BBBB 16.890 13.076 1.292
## blockMN1:LabelE 47.002 12.114 3.880
## BBBB:LabelE -33.098 9.727 -3.403
## blockMN1:BBBB:LabelE -49.074 17.135 -2.864
##
## Correlation of Fixed Effects:
## (Intr) blcMN1 BBBB LabelE gendrM blMN1:BBBB bMN1:L BBBB:L
## blockMN1 -0.188
## BBBB -0.618 0.152
## LabelE -0.391 0.238 0.124
## genderM -0.239 0.017 -0.031 0.003
## blcMN1:BBBB 0.129 -0.705 -0.207 -0.168 0.003
## blckMN1:LbE 0.140 -0.662 -0.116 -0.359 -0.003 0.467
## BBBB:LabelE 0.174 -0.276 -0.278 -0.448 -0.002 0.373 0.417
## bMN1:BBBB:L -0.099 0.468 0.158 0.254 0.001 -0.661 -0.707 -0.568Interpretation:
AF model:
blockAF1 10.329 7.846 1.316
in NB group, the F1 (of female speakers) in the AF test condition did not differ from pretest
BBBB 5.255 16.573 0.317
The F1 in /I/ in BB group at pretest does not differ from NB group among female speakers
LabelE 167.312 13.098 12.774
the F1 in /E/ is 167.312 Hz higher than /I/ in the females speakers in the NB group at pretest
genderM -93.417 21.272 -4.392
Male speakers in the NB group have an F1 that is 93.417Hz lower than the female speakers in the NB group at pretest
blockAF1:BBBB 31.225 10.767 2.900
The increase of F1 from pretest to AF1 is 31.225 Hz bigger in the BB group (for female speakers). Since BB group did not differ from NB group at pretest, we can conclude that having a bite block increased the F1 of 31.225 Hz in the AF1 block for female speakers.
blockAF1:LabelE -27.850 10.317 -2.699
The increase of F1 from /I/ to /E/ is 27.850 Hz smaller in the AF (and no bite block, female) group at test block compared to pretest; i.e., the /I/ -/E/ contrast is 27.850 Hz smaller in the AF group at test –> practice effect?
BBBB:LabelE -33.366 8.280 -4.030
At pretest, the increase of F1 from /I/ to /E/ is 33. 366 Hz smaller in the BB group than the NB group. BB group behaves differently at pretest again. But since the F1 in /I/ did not differ between BB and NB group (see result BBBB where t = 0.317), it’s mainly the /E/ that differed between NB and BB group at pretest.
MN model:
blockMN1 53.786 9.226 5.830
For female speakers in the NB group, being in the MN condition increased (from pretest) the F1 of /I/ for 53.786 Hz -> Lombard
BBBB 4.803 17.433 0.276
For female speakers at pretest, the F1 in /I/ does not differ between NB and BB group
LabelE 167.281 11.276 14.835
For female speakers in the NB group, at pretest, the F1 in /E/ is 167.281 Hz higher than /I/
genderM -93.159 22.364 -4.166
At pretest, in the NB group, for vowel /I/, male speakers have an F1 that is 93.159 Hz lower than the female speakers in the same group at the same stage of testing.
blockMN1:BBBB 16.890 13.076 1.292
The increase of F1 from pretest to MN1 (females, vowel /I/) does not differ between BB and NB groups.
blockMN1:LabelE 47.002 12.114 3.880
The increase of F1 from /I/ to /E/ from pretest to MN1 (i.e., the F1 difference between /I/ and /E/ from pretest to MN) is 47.002 Hz bigger –> hyperarticulation in Lombard?
blockMN1:BBBB:LabelE -49.074 17.135 -2.864
The difference between /I/ and /E/ between BB and NB group (the difference of the size of the /I/- /E/ contrast between BB and NB group) is -49.074 Hz at MN1 when compared to pretest. That the BB group has a 49.074 Hz smaller /I/ -/E/ contrast change (pretest to MN1) than the NB group.
#code male as reference level to check whether the BB group have a different F1 at AF1
# (the alternative would be to code the contrast for gender as -0.5/+0.5, which then has the mean of F and M as the reference level. But literature has shown a gender difference in tongue root mobility, so we have reason to distinguish the two)
af$gender <- as.factor(af$gender)
af$gender <- relevel(af$gender, ref = "M")
mn$gender <- as.factor(mn$gender)
mn$gender <- relevel(mn$gender, ref = "M")
f1hzafm <- lmer(F1Hz ~ block * BB * block* Label + gender + (1|subj) + (1|word), data = af)
summary(f1hzafm)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Hz ~ block * BB * block * Label + gender + (1 | subj) + (1 |
## word)
## Data: af
##
## REML criterion at convergence: 16221.6
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.2819 -0.5589 0.0282 0.5470 11.0900
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 3238 56.90
## word (Intercept) 204 14.28
## Residual 4069 63.79
## Number of obs: 1445, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 347.435 23.110 15.034
## blockAF1 10.329 7.846 1.316
## BBBB 5.255 16.573 0.317
## LabelE 167.312 13.098 12.774
## genderF 93.417 21.272 4.392
## blockAF1:BBBB 31.225 10.767 2.900
## blockAF1:LabelE -27.850 10.317 -2.699
## BBBB:LabelE -33.366 8.280 -4.030
## blockAF1:BBBB:LabelE 2.991 14.189 0.211
##
## Correlation of Fixed Effects:
## (Intr) blcAF1 BBBB LabelE gendrF blAF1:BBBB bAF1:L BBBB:L
## blockAF1 -0.107
## BBBB -0.396 0.136
## LabelE -0.282 0.170 0.082
## genderF -0.779 0.013 0.031 -0.001
## blcAF1:BBBB 0.069 -0.729 -0.191 -0.124 0.003
## blckAF1:LbE 0.072 -0.657 -0.103 -0.263 0.003 0.479
## BBBB:LabelE 0.091 -0.269 -0.249 -0.328 0.001 0.381 0.417
## bAF1:BBBB:L -0.052 0.478 0.145 0.192 -0.003 -0.660 -0.727 -0.584f1hzmnm <- lmer(F1Hz ~ block * BB * block* Label + gender + (1|subj) + (1|word), data = mn)
summary(f1hzmnm)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Hz ~ block * BB * block * Label + gender + (1 | subj) + (1 |
## word)
## Data: mn
##
## REML criterion at convergence: 16195.6
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -4.7920 -0.4615 0.0561 0.4755 9.1062
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 3457.4 58.80
## word (Intercept) 117.1 10.82
## Residual 5611.9 74.91
## Number of obs: 1404, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 347.805 23.545 14.772
## blockMN1 53.786 9.226 5.830
## BBBB 4.803 17.433 0.276
## LabelE 167.281 11.276 14.835
## genderF 93.159 22.364 4.166
## blockMN1:BBBB 16.890 13.076 1.292
## blockMN1:LabelE 47.002 12.114 3.880
## BBBB:LabelE -33.098 9.727 -3.403
## blockMN1:BBBB:LabelE -49.074 17.135 -2.864
##
## Correlation of Fixed Effects:
## (Intr) blcMN1 BBBB LabelE gendrF blMN1:BBBB bMN1:L BBBB:L
## blockMN1 -0.099
## BBBB -0.408 0.152
## LabelE -0.237 0.238 0.124
## genderF -0.804 -0.017 0.031 -0.003
## blcMN1:BBBB 0.082 -0.705 -0.207 -0.168 -0.003
## blckMN1:LbE 0.083 -0.662 -0.116 -0.359 0.003 0.467
## BBBB:LabelE 0.105 -0.276 -0.278 -0.448 0.002 0.373 0.417
## bMN1:BBBB:L -0.060 0.468 0.158 0.254 -0.001 -0.661 -0.707 -0.568same results except intercept.
#model plot
data_summary <- function(data, varname, groupnames){
require(plyr)
summary_func <- function(x, col){
c(mean = mean(x[[col]], na.rm=TRUE),
sd = sd(x[[col]], na.rm=TRUE))
}
data_sum<-ddply(data, groupnames, .fun=summary_func,
varname)
data_sum <- rename(data_sum, c("mean" = varname))
return(data_sum)
}
library(ggplot2)
f1i <- af |> filter(Label == "I")
f1e <- af |> filter(Label == "E")
f1ierr <- data_summary(f1i, varname="F1Hz",
groupnames=c("block", "BB"))## Loading required package: plyr## ------------------------------------------------------------------------------## You have loaded plyr after dplyr - this is likely to cause problems.
## If you need functions from both plyr and dplyr, please load plyr first, then dplyr:
## library(plyr); library(dplyr)## ------------------------------------------------------------------------------##
## Attaching package: 'plyr'## The following objects are masked from 'package:dplyr':
##
## arrange, count, desc, failwith, id, mutate, rename, summarise,
## summarizef1eerr <- data_summary(f1e, varname="F1Hz",
groupnames=c("block", "BB"))
p1<- ggplot(data=f1i, aes(x=BB, y=F1Hz, fill=block)) +
geom_bar(stat="identity", position=position_dodge())+
labs (title = "F1 of /I/ at pre test and test, in AF condition")
p1 + geom_errorbar(data = f1ierr, aes(ymin=F1Hz-sd, ymax=F1Hz+sd), width=.2,
position=position_dodge(.9))
p2 <- ggplot(data=f1e, aes(x=BB, y=F1Hz, fill=block)) +
geom_bar(stat="identity", position=position_dodge())+
labs (title = "F1 of /E/ at pre test and test, in AF condition")
p2 + geom_errorbar(data = f1eerr, aes(ymin=F1Hz-sd, ymax=F1Hz+sd), width=.2,
position=position_dodge(.9))
#model plot
library(ggplot2)
mni <- mn |> filter(Label == "I")
ggplot(data=mni, aes(x=BB, y=F1Hz, fill=block)) +
geom_bar(stat="identity", position=position_dodge())+
labs (title = "F1 of /I/ at pre test and test, in MN condition")
mne <- mn |> filter(Label == "E")
ggplot(data=mne, aes(x=BB, y=F1Hz, fill=block)) +
geom_bar(stat="identity", position=position_dodge())+
labs (title = "F1 of /E/ at pre test and test, in MN condition")
#add in error bars
data_summary <- function(data, varname, groupnames){
require(plyr)
summary_func <- function(x, col){
c(mean = mean(x[[col]], na.rm=TRUE),
sd = sd(x[[col]], na.rm=TRUE))
}
data_sum<-ddply(data, groupnames, .fun=summary_func,
varname)
data_sum <- rename(data_sum, c("mean" = varname))
return(data_sum)
}
f1ierr <- data_summary(f1i, varname="F1Hz",
groupnames=c("block", "BB"))
f1eerr <- data_summary(f1e, varname="F1Hz",
groupnames=c("block", "BB"))
head(f1ierr)## block BB F1Hz sd
## 1 Pre NB 425.7174 62.73284
## 2 Pre BB 429.4696 62.73190
## 3 AF1 NB 430.3596 60.44468
## 4 AF1 BB 470.5115 101.61234p <- ggplot(data=f1ierr, aes(x=BB, y=F1Hz, fill=block)) +
geom_bar(stat="identity", position=position_dodge()) +
geom_errorbar(aes(ymin=F1Hz-sd, ymax=F1Hz+sd), width=.2,
position=position_dodge(.9))
p + scale_fill_brewer(palette="Paired") + theme_minimal()+labs (title = "F1 of /I/ at pre test and test, in AF condition")
library(effects)
plot(effect( "block * BB * Label", f1hzmn))
plot(effect( "block * BB * Label", f1hzaf))
#f2 models
af$gender <- as.factor(af$gender)
af$gender <- relevel(af$gender, ref = "F")
f2hzaf <- lmer(F2Hz ~ block * BB * block* Label + gender + (1|subj) + (1|word), data = af)
summary(f2hzaf)## Linear mixed model fit by REML ['lmerMod']
## Formula: F2Hz ~ block * BB * block * Label + gender + (1 | subj) + (1 |
## word)
## Data: af
##
## REML criterion at convergence: 18372.4
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.3025 -0.5015 -0.0194 0.4816 5.5072
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 13750 117.3
## word (Intercept) 10159 100.8
## Residual 18103 134.5
## Number of obs: 1445, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 2202.441 63.558 34.653
## blockAF1 -7.619 16.547 -0.460
## BBBB -34.253 34.252 -1.000
## LabelE -252.224 83.252 -3.030
## genderM -238.697 43.881 -5.440
## blockAF1:BBBB -669.195 22.706 -29.472
## blockAF1:LabelE -19.466 21.763 -0.894
## BBBB:LabelE 35.904 17.466 2.056
## blockAF1:BBBB:LabelE 221.029 29.929 7.385
##
## Correlation of Fixed Effects:
## (Intr) blcAF1 BBBB LabelE gendrM blAF1:BBBB bAF1:L BBBB:L
## blockAF1 -0.073
## BBBB -0.276 0.139
## LabelE -0.655 0.056 0.028
## genderM -0.106 -0.013 -0.031 0.000
## blcAF1:BBBB 0.055 -0.729 -0.195 -0.041 -0.003
## blckAF1:LbE 0.057 -0.657 -0.106 -0.087 -0.003 0.479
## BBBB:LabelE 0.071 -0.269 -0.254 -0.109 -0.001 0.381 0.417
## bAF1:BBBB:L -0.042 0.478 0.148 0.064 0.003 -0.660 -0.727 -0.584f2hzmn <- lmer(F2Hz ~ block * BB * block* Label + gender + (1|subj) + (1|word), data = mn)
summary(f2hzmn)## Linear mixed model fit by REML ['lmerMod']
## Formula: F2Hz ~ block * BB * block * Label + gender + (1 | subj) + (1 |
## word)
## Data: mn
##
## REML criterion at convergence: 17676.1
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.3292 -0.5705 -0.0235 0.5351 5.5431
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 11445 106.98
## word (Intercept) 9678 98.38
## Residual 15989 126.45
## Number of obs: 1404, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 1954.14 70.06 27.894
## blockMN1 -64.91 15.58 -4.165
## BBBB -33.97 31.37 -1.083
## LabelE -252.46 81.19 -3.110
## genderF 250.16 40.47 6.181
## blockMN1:BBBB -720.16 22.09 -32.603
## blockMN1:LabelE 23.59 20.45 1.154
## BBBB:LabelE 35.95 16.42 2.190
## blockMN1:BBBB:LabelE 168.52 28.92 5.827
##
## Correlation of Fixed Effects:
## (Intr) blcMN1 BBBB LabelE gendrF blMN1:BBBB bMN1:L BBBB:L
## blockMN1 -0.056
## BBBB -0.247 0.142
## LabelE -0.579 0.056 0.027
## genderF -0.489 -0.016 0.031 -0.001
## blcMN1:BBBB 0.046 -0.705 -0.194 -0.039 -0.003
## blckMN1:LbE 0.047 -0.662 -0.108 -0.084 0.002 0.467
## BBBB:LabelE 0.059 -0.276 -0.261 -0.105 0.002 0.373 0.417
## bMN1:BBBB:L -0.034 0.468 0.148 0.060 -0.001 -0.660 -0.707 -0.568plot(effect( "block * BB * Label", f1hzaf))
#prepost model
#splitting up MN1 and AF1, compare with pretest respectively
prepost <- db |> filter(block == "Pre" | block == "Post")
#contrast coding
prepost$BB <- as.factor(prepost$BB)
prepost$BB <- relevel(prepost$BB, ref = "NB")
prepost$Label <- as.factor(prepost$Label)
prepost$Label <- relevel(prepost$Label, ref = "I")
prepost$block <- as.factor(prepost$block)
prepost$block <- relevel(prepost$block, ref = "Pre")
prepost$gender <- as.factor(prepost$gender)
prepost$gender <- relevel(prepost$gender, ref = "F")
f1hzpp <- lmer(F1Hz ~ block * BB * Label + gender + (1|subj) + (1|word), data = prepost)
summary(f1hzpp)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Hz ~ block * BB * Label + gender + (1 | subj) + (1 | word)
## Data: prepost
##
## REML criterion at convergence: 21501.3
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.6421 -0.5144 0.0187 0.5142 18.3661
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 2348 48.46
## word (Intercept) 284 16.85
## Residual 4483 66.95
## Number of obs: 1902, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 439.381 14.571 30.154
## blockPost 1.010 6.224 0.162
## BBBB 5.049 14.561 0.347
## LabelE 167.737 15.117 11.096
## genderM -84.508 18.188 -4.646
## blockPost:BBBB -6.765 8.681 -0.779
## blockPost:LabelE -5.662 8.832 -0.641
## BBBB:LabelE -33.768 8.690 -3.886
## blockPost:BBBB:LabelE 13.845 12.293 1.126
##
## Correlation of Fixed Effects:
## (Intr) blckPs BBBB LabelE gendrM blP:BBBB blP:LE BBBB:L
## blockPost -0.215
## BBBB -0.512 0.215
## LabelE -0.518 0.207 0.089
## genderM -0.192 0.000 -0.030 0.001
## blckPs:BBBB 0.154 -0.717 -0.298 -0.149 -0.001
## blckPst:LbE 0.152 -0.705 -0.152 -0.293 -0.002 0.505
## BBBB:LabelE 0.154 -0.360 -0.298 -0.298 -0.001 0.499 0.510
## blP:BBBB:LE -0.109 0.506 0.210 0.211 0.001 -0.706 -0.719 -0.707f2hzpp <- lmer(F2Hz ~ block * BB * Label + gender + (1|subj) + (1|word), data = prepost)
summary(f2hzpp)## Linear mixed model fit by REML ['lmerMod']
## Formula: F2Hz ~ block * BB * Label + gender + (1 | subj) + (1 | word)
## Data: prepost
##
## REML criterion at convergence: 23423.6
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -6.2328 -0.5602 -0.0181 0.5161 6.4825
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 13382 115.7
## word (Intercept) 14516 120.5
## Residual 12062 109.8
## Number of obs: 1902, groups: subj, 54; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 2204.760 73.819 29.867
## blockPost -33.127 10.210 -3.245
## BBBB -33.930 33.092 -1.025
## LabelE -252.803 98.908 -2.556
## genderM -254.596 42.814 -5.947
## blockPost:BBBB -2.707 14.240 -0.190
## blockPost:LabelE 26.317 14.489 1.816
## BBBB:LabelE 37.231 14.256 2.612
## blockPost:BBBB:LabelE -24.773 20.165 -1.229
##
## Correlation of Fixed Effects:
## (Intr) blckPs BBBB LabelE gendrM blP:BBBB blP:LE BBBB:L
## blockPost -0.070
## BBBB -0.229 0.155
## LabelE -0.670 0.052 0.016
## genderM -0.089 0.000 -0.032 0.000
## blckPs:BBBB 0.050 -0.717 -0.215 -0.037 0.000
## blckPst:LbE 0.049 -0.705 -0.109 -0.074 -0.001 0.505
## BBBB:LabelE 0.050 -0.360 -0.215 -0.075 -0.001 0.499 0.510
## blP:BBBB:LE -0.035 0.506 0.152 0.053 0.001 -0.706 -0.718 -0.707#plotting the one significant result
plot(effect( "BB * Label", f1hzpp))## NOTE: BB:Label is not a high-order term in the model
mn2subj <- c("2", "3", "5", "6", "8", "9", "10", "12", "15", "17", "19", "26", "30", "31", "33", "35", "38", "40", "42", "43", "46", "48", "50", "51", "52", "53", "54", "56", "58", "60")
ppaf <- db |> filter(subj %in% mn2subj)|> filter(block == "Pre" | block == "Post")
ppmn <- db |> filter(!subj %in% mn2subj)|> filter(block == "Pre" | block == "Post")
#change df name
ppaf$BB <- as.factor(ppaf$BB)
ppaf$BB <- relevel(ppaf$BB, ref = "NB")
ppaf$Label <- as.factor(ppaf$Label)
ppaf$Label <- relevel(ppaf$Label, ref = "I")
ppaf$block <- as.factor(ppaf$block)
ppaf$block <- relevel(ppaf$block, ref = "Pre")
ppaf$gender <- as.factor(ppaf$gender)
ppaf$gender <- relevel(ppaf$gender, ref = "F")
ppmn$BB <- as.factor(ppmn$BB)
ppmn$BB <- relevel(ppmn$BB, ref = "NB")
ppmn$Label <- as.factor(ppmn$Label)
ppmn$Label <- relevel(ppmn$Label, ref = "I")
ppmn$block <- as.factor(ppmn$block)
ppmn$block <- relevel(ppmn$block, ref = "Pre")
ppmn$gender <- as.factor(ppmn$gender)
ppmn$gender <- relevel(ppmn$gender, ref = "F")
#run lmers
ppmnf1 <- lmer(F1Hz ~ block * BB * Label + gender + (1|subj) + (1|word), data = ppmn)
summary(ppmnf1)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Hz ~ block * BB * Label + gender + (1 | subj) + (1 | word)
## Data: ppmn
##
## REML criterion at convergence: 10241.8
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.1742 -0.6261 0.0390 0.6322 4.7036
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 2253.4 47.47
## word (Intercept) 279.2 16.71
## Residual 4004.9 63.28
## Number of obs: 918, groups: subj, 26; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 429.0504 17.5832 24.401
## blockPost -13.0507 8.3471 -1.563
## BBBB -0.2766 20.4050 -0.014
## LabelE 187.0542 15.9947 11.695
## genderM -60.4472 35.7904 -1.689
## blockPost:BBBB 11.3719 11.8186 0.962
## blockPost:LabelE 8.4734 11.8182 0.717
## BBBB:LabelE -44.9571 11.7788 -3.817
## blockPost:BBBB:LabelE -4.6205 16.7135 -0.276
##
## Correlation of Fixed Effects:
## (Intr) blckPs BBBB LabelE gendrM blP:BBBB blP:LE BBBB:L
## blockPost -0.240
## BBBB -0.581 0.206
## LabelE -0.456 0.263 0.108
## genderM -0.157 0.000 0.000 0.000
## blckPs:BBBB 0.169 -0.706 -0.289 -0.186 -0.001
## blckPst:LbE 0.169 -0.706 -0.146 -0.369 -0.001 0.499
## BBBB:LabelE 0.170 -0.357 -0.290 -0.370 0.000 0.500 0.500
## blP:BBBB:LE -0.120 0.499 0.204 0.261 0.000 -0.707 -0.707 -0.705ppmnf2 <- lmer(F2Hz ~ block * BB * Label + gender + (1|subj) + (1|word), data = ppmn)
summary(ppmnf2)## Linear mixed model fit by REML ['lmerMod']
## Formula: F2Hz ~ block * BB * Label + gender + (1 | subj) + (1 | word)
## Data: ppmn
##
## REML criterion at convergence: 11344.2
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.7871 -0.5714 -0.0224 0.5501 5.2096
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 16284 127.6
## word (Intercept) 15379 124.0
## Residual 13048 114.2
## Number of obs: 918, groups: subj, 26; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 2196.729 80.914 27.149
## blockPost -9.326 15.067 -0.619
## BBBB 18.426 52.271 0.353
## LabelE -248.993 102.372 -2.432
## genderM -134.288 94.958 -1.414
## blockPost:BBBB -33.548 21.333 -1.573
## blockPost:LabelE 8.859 21.332 0.415
## BBBB:LabelE 3.701 21.261 0.174
## blockPost:BBBB:LabelE -13.084 30.168 -0.434
##
## Correlation of Fixed Effects:
## (Intr) blckPs BBBB LabelE gendrM blP:BBBB blP:LE BBBB:L
## blockPost -0.094
## BBBB -0.323 0.145
## LabelE -0.633 0.074 0.021
## genderM -0.090 0.000 0.000 0.000
## blckPs:BBBB 0.066 -0.706 -0.204 -0.052 0.000
## blckPst:LbE 0.066 -0.706 -0.103 -0.104 0.000 0.499
## BBBB:LabelE 0.067 -0.357 -0.204 -0.104 0.000 0.500 0.500
## blP:BBBB:LE -0.047 0.499 0.144 0.073 0.000 -0.707 -0.707 -0.705ppaff1 <- lmer(F1Hz ~ block * BB * Label + gender + (1|subj) + (1|word), data = ppaf)
summary(ppaff1)## Linear mixed model fit by REML ['lmerMod']
## Formula: F1Hz ~ block * BB * Label + gender + (1 | subj) + (1 | word)
## Data: ppaf
##
## REML criterion at convergence: 11150.9
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -3.5734 -0.4274 -0.0041 0.4300 17.4702
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 2568.4 50.68
## word (Intercept) 279.3 16.71
## Residual 4758.7 68.98
## Number of obs: 984, groups: subj, 28; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 450.991 18.954 23.794
## blockPost 14.974 9.039 1.657
## BBBB 9.194 21.138 0.435
## LabelE 147.979 16.427 9.008
## genderM -98.369 22.729 -4.328
## blockPost:BBBB -24.441 12.438 -1.965
## blockPost:LabelE -18.871 12.858 -1.468
## BBBB:LabelE -21.214 12.506 -1.696
## blockPost:BBBB:LabelE 31.051 17.638 1.760
##
## Correlation of Fixed Effects:
## (Intr) blckPs BBBB LabelE gendrM blP:BBBB blP:LE BBBB:L
## blockPost -0.240
## BBBB -0.585 0.215
## LabelE -0.431 0.276 0.118
## genderM -0.277 0.000 -0.038 0.002
## blckPs:BBBB 0.174 -0.727 -0.294 -0.201 0.000
## blckPst:LbE 0.169 -0.703 -0.151 -0.396 -0.002 0.511
## BBBB:LabelE 0.174 -0.363 -0.293 -0.407 -0.002 0.498 0.520
## blP:BBBB:LE -0.123 0.512 0.208 0.289 0.001 -0.705 -0.729 -0.709ppaff2 <- lmer(F2Hz ~ block * BB * Label + gender + (1|subj) + (1|word), data = ppaf)
summary(ppaff2)## Linear mixed model fit by REML ['lmerMod']
## Formula: F2Hz ~ block * BB * Label + gender + (1 | subj) + (1 | word)
## Data: ppaf
##
## REML criterion at convergence: 11982.5
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.0531 -0.5362 -0.0269 0.4896 7.0453
##
## Random effects:
## Groups Name Variance Std.Dev.
## subj (Intercept) 9952 99.76
## word (Intercept) 13724 117.15
## Residual 10888 104.35
## Number of obs: 984, groups: subj, 28; word, 6
##
## Fixed effects:
## Estimate Std. Error t value
## (Intercept) 2208.41 74.42 29.674
## blockPost -56.61 13.67 -4.140
## BBBB -78.52 40.11 -1.958
## LabelE -256.12 96.65 -2.650
## genderM -276.74 44.27 -6.251
## blockPost:BBBB 27.38 18.81 1.456
## blockPost:LabelE 43.20 19.45 2.221
## BBBB:LabelE 66.89 18.92 3.536
## blockPost:BBBB:LabelE -37.25 26.68 -1.396
##
## Correlation of Fixed Effects:
## (Intr) blckPs BBBB LabelE gendrM blP:BBBB blP:LE BBBB:L
## blockPost -0.092
## BBBB -0.283 0.171
## LabelE -0.649 0.071 0.024
## genderM -0.137 0.000 -0.039 0.000
## blckPs:BBBB 0.067 -0.727 -0.235 -0.052 0.000
## blckPst:LbE 0.065 -0.703 -0.120 -0.102 -0.002 0.511
## BBBB:LabelE 0.067 -0.363 -0.233 -0.105 -0.001 0.498 0.520
## blP:BBBB:LE -0.047 0.512 0.165 0.074 0.001 -0.705 -0.729 -0.709